# Consensus > This file is very large and may be truncated. For an index of this content, please refer to https://docs.consensus.app/llms.txt instead. > Developer documentation for Consensus, the AI research engine. # Consensus API compared with PubMed, Europe PMC, OpenAlex and Semantic Scholar Source: https://docs.consensus.app/api-comparison Why teams build on the Consensus API, how it works alongside PubMed, Europe PMC, OpenAlex, Semantic Scholar and Crossref, and how the APIs compare on coverage, search, evidence metadata, full text, pricing and rate limits. The Consensus API searches the scholarly literature by meaning and returns papers with the evidence details already attached. Its research agent goes further: it crawls citations, traces authors and plans multi-step searches, the way a researcher would. This page explains what it's built for, how teams use it alongside the free scholarly APIs (PubMed, Europe PMC, OpenAlex, Semantic Scholar and Crossref), and how they compare. The pages under **Compare scholarly APIs** go deeper on each topic. ## Why teams choose the Consensus API > [card] [400M+ papers across every field](/get-started/where-papers-come-from) > > 400 million scholarly articles, 350 million of them peer-reviewed, from astronomy to sociology. Includes all of PubMed. Updated weekly. > [card] [Full text from publisher partners](/get-started/full-text-access) > > Search reaches into the full text of open-access papers and of paywalled papers from 7 of the 10 largest academic publishers. > [card] [Ask in plain language](/get-started/how-search-works) > > Hybrid semantic and keyword search, ranked for scientific relevance. Send the question as written. Boolean queries work too. > [card] [First-party evidence filters](/api-reference/search) > > Filter by study type, sample size, journal quartile, human or animal studies, country of study, study duration, citations and more. > [card] [Full-text snippets for grounding](/api-full-text) > > Get the passages from each paper that answer your query, labeled by section, ready to put in front of an LLM. > [card] [A research agent, not just search](/api-quickstart-threads) > > The Threads API plans a search strategy, crawls citations, looks up authors and reads full papers, then writes an answer with numbered citations and the quotes behind them. Publisher partners whose full text Consensus searches include: ![Wiley](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/wiley.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=89f93113e74bb0090ce94a5b4f55c24a) ![Taylor & Francis](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/taylor-and-francis.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=2075ac1161cb69642ac53438254b2aae) ![SAGE](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/sage.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=da227aff27e194e0395d10dddb992ece) ![Wolters Kluwer](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/wolters-kluwer.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=8d67fdc3b4fa013aa8cc8429b49de5f7) ![Oxford University Press](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/oxford-university-press.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=4159326ee81748817a0c41594c71c17d) ![American Chemical Society](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/acs.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=6b8e12edf4b83daef036c2aa4b8c14a9) ![De Gruyter Brill](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/de-gruyter-brill.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=efed8f09bb4ea863bfa97cc33566c933) ![BMJ](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/bmj.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=73f5754cbec30c1ad3ef2ce20a71d81c) ![American Psychological Association](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/apa.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=29cfd1c4ea69e51519e5a8edda1081a5) ![AAAS](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/aaas.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=6ebd3cc659c6f52f1f4e95d8af2b8aec) ![JAMA Network](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/jama.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=d9a7d8740a81df20c319564a003723bd) ![American Society for Microbiology](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/asm.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=b0ea127517100a30f2a38ca56eb60467) Partner full text improves which papers rank and powers analysis in the Consensus app. The snippets the API returns come from open-access papers. See [Full text in the API](/api-full-text). Every paper in a response carries the details you'd otherwise have to dig for: `study_type`, `sample_size`, `sjr_best_quartile`, `population_type`, `countries_of_study`, `citation_count`, `is_preprint` and a one-sentence `takeaway`. Your code, or your model, can weigh the evidence without opening the paper. ## How teams combine Consensus with open APIs Most teams don't pick one API. They use Consensus to find the evidence, then call the free APIs for what those do best. **1. Find and rank the papers with Consensus** Send the research question as written. Filter to the study designs and sample sizes you trust, and pull full-text snippets for the papers your model will read. **2. Enrich the results with open metadata** Every Consensus result has a DOI. Use it to pull citation graphs and similar papers from Semantic Scholar, author, institution and funder data from OpenAlex, or registration metadata from Crossref. **3. Run a Boolean strategy where a protocol requires one** For the search you report in a systematic review, run a reproducible Boolean query in PubMed or Europe PMC, and use Consensus to check you haven't missed the strongest studies. This works smoothly because Consensus builds on the same open records: its index aggregates Semantic Scholar, OpenAlex and PubMed, then adds its own crawl of the scholarly web, publisher full text, evidence classification and ranking. ## At a glance A dash (—) means we didn't find the feature in that provider's API documentation as of November 2026. ### Coverage and search | | Coverage | How you search | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | | **OpenAlex** | | | | **Semantic Scholar** | | | | **Crossref** | | | ### Evidence metadata and full text | | Evidence metadata per paper | Full text | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | | **OpenAlex** | | | | **Semantic Scholar** | | | | **Crossref** | | | ### Pricing and rate limits | | Pricing | Rate limits | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | | **OpenAlex** | | | | **Semantic Scholar** | | | | **Crossref** | | | A Consensus call covers up to 100 papers, and the API and MCP server share one monthly pool. See [plans and access](/api-plans-and-access). ## Compare in depth > [card] [Biomedical search: Consensus, PubMed and Europe PMC](/api-comparison/pubmed-europe-pmc) > > Semantic search over PubMed's records, study-design filters, and when a Boolean strategy is the right tool. > [card] [Open metadata: OpenAlex, Semantic Scholar and Crossref](/api-comparison/openalex-semantic-scholar-crossref) > > What each open API does best, and how to enrich Consensus results with it. > [card] [Retrieval for RAG and AI agents](/api-comparison/rag-and-ai-agents) > > What a RAG pipeline needs from a search API, and when web search APIs like Exa and Tavily fit. > [card] [Pricing, rate limits and data privacy](/api-comparison/pricing-and-privacy) > > Costs, limits and interfaces side by side, and what each provider does with your queries. > [card] [Search quality evaluations](/api-comparison/evaluations) > > Two blind relevance studies: more relevant than Google Scholar, and more relevant than GPT-5.5 and Opus 4.8 in a fraction of the time. ## FAQ > [accordion] Is Consensus built on PubMed, OpenAlex and Semantic Scholar? > > Partly. Consensus aggregates metadata and abstracts from Semantic Scholar, OpenAlex and PubMed, crawls the scholarly web to fill the gaps, and adds full text from open-access papers and publisher partners. On top of that it classifies each paper's study design and sample size and ranks results for scientific relevance. See [where the papers come from](/get-started/where-papers-come-from). > [accordion] Is there a free Consensus API? > > Yes. Every plan includes API access. The Free plan has 30 calls a month and up to 20 papers per request, without pagination or full-text snippets. See [plans and access](/api-plans-and-access). > [accordion] Does the Consensus API return full text? > > Every search matches against full text where it's available. On paid plans, each search also returns up to 3 query-relevant passages per paper from open-access papers. See [Full text in the API](/api-full-text). > [accordion] Does Consensus train AI models on my queries? > > No. Customer data isn't used to train AI models, and the LLM providers Consensus uses keep no data from requests. See [Pricing, rate limits and data privacy](/api-comparison/pricing-and-privacy). > [accordion] What's the difference between the Consensus API and the Consensus MCP server? > > The API is a REST interface for your own code. The [MCP server](/consensus-mcp) gives AI assistants and agents, such as Claude, ChatGPT and Cursor, the same search and research agent as tools. Both draw on one monthly pool of calls. > [accordion] Is Consensus related to goconsensus.com? > > No. goconsensus.com is an unrelated sales-demo software company that also publishes a "Consensus MCP". Consensus, the AI search engine for research, is at [consensus.app](https://consensus.app), with its API at `api.consensus.app` and its MCP server at `mcp.consensus.app`. ## Sources Facts about other providers were checked against their official documentation on 5 November 2026. Limits and prices change, so check the source before you build on a number. 1. [About PubMed](https://pubmed.ncbi.nlm.nih.gov/about/): "more than 40 million citations and abstracts of biomedical literature"; "does not include full text journal articles"; a free resource. 2. [PubMed User Guide](https://pubmed.ncbi.nlm.nih.gov/help/): Boolean operators, field tags, MeSH and Best Match sorting. 3. [E-utilities in depth, ESearch](https://www.ncbi.nlm.nih.gov/books/NBK25499/): `sort=relevance` is the default Best Match order on web PubMed. 4. [About Europe PMC](https://europepmc.org/About): 49.2 million abstracts and 12.5 million full-text articles; searches abstracts and full text; searches can be limited to clinical trials, preprints or article sections such as Methods; free to anyone. 5. [OpenAlex works](https://help.openalex.org/data/works/): more than 320 million works in the core catalog. 6. [OpenAlex search](https://help.openalex.org/api/searching/): `search=` matches works by title, abstract, full text and keywords; Boolean and phrase searches are supported. 7. [OpenAlex semantic search](https://help.openalex.org/api/semantic-search/): embeds title, abstract and venue name; up to 50 results per query; 1 request a second; 17 supported filters, and others such as `cited_by_count` return a 400 error. 8. [Semantic Scholar API overview](https://www.semanticscholar.org/product/api): 214 million papers; unauthenticated requests share 1,000 requests a second across all users and may be throttled under heavy use; the introductory limit for an API key is 1 request a second. The API also serves SPECTER2 embeddings and a Recommendations service for similar papers. 9. [Semantic Scholar Academic Graph API reference](https://api.semanticscholar.org/api-docs/graph): `/paper/search` (up to 1,000 relevance-ranked results), `/paper/search/bulk` (Boolean), `/snippet/search` (passages of about 500 words from title, abstract and body), and the `publicationTypes`, `minCitationCount` and `venue` filters. 10. `total-results` from `https://api.crossref.org/works?rows=0` on 5 November 2026: 188,499,975. 11. [Crossref REST API](https://www.crossref.org/documentation/retrieve-metadata/rest-api/): bibliographic metadata including funding, licence and abstracts; no sign-up required; most metadata isn't subject to copyright. 12. [NLM publication types](https://www.nlm.nih.gov/mesh/pubtypes.html). 13. [About PMC](https://pmc.ncbi.nlm.nih.gov/about/intro/): "more than 10 million full-text article records". 14. [Europe PMC developer resources](https://europepmc.org/developers): REST, SOAP, OAI-PMH and FTP access to open-access content and metadata; automated bulk download of other content isn't allowed. The [Articles RESTful API](https://europepmc.org/RestfulWebService) includes `/{id}/fullTextXML`. 15. [OpenAlex study designs](https://help.openalex.org/data/study-designs/): seven designs tagged automatically from title and abstract; each value right at least 96% of the time. 16. [OpenAlex full text](https://help.openalex.org/access/fulltext/): cached PDF and TEI XML for more than 60 million open-access works. 17. [A General Introduction to the E-utilities](https://www.ncbi.nlm.nih.gov/books/NBK25497/): more than 3 requests a second without an API key returns an error; with a key, up to 10 a second by default; higher rates by request. Requests without registered `tool` and `email` values may be blocked. 18. Europe PMC team on the [Europe PMC web services forum](https://groups.google.com/a/ebi.ac.uk/g/epmc-webservices/c/x3PLL09mels), 1 October 2024: the rate limit is 10 requests a second (500 a minute), and no API key is required. 19. [OpenAlex pricing](https://help.openalex.org/access/pricing/): all data CC0; \$1 of free API usage a day with a key; annual plans from \$5,000 a year; the whole database can be downloaded as a snapshot refreshed every quarter. A keyless response notes that it shares a \$0.10 daily budget per IP address. 20. [OpenAlex authentication and rate limits](https://help.openalex.org/api/authentication/): 10 requests a second with no key, 30 with a free key or prepaid usage, 50 on an annual plan; semantic search is limited to 1 a second. 21. [Crossref access and authentication](https://www.crossref.org/documentation/retrieve-metadata/rest-api/access-and-authentication/): public pool 5 requests a second, polite pool 10, Metadata Plus 150. # Search quality evaluations Source: https://docs.consensus.app/api-comparison/evaluations Two blind relevance evaluations of Consensus search: against Google Scholar on 500 real queries, and against GPT-5.5 and Opus 4.8 with web search on 100 real queries. Methods, results and caveats. Consensus has published two blind evaluations of how relevant its search results are. Both used real, anonymized user queries and independent PhD-level judges who didn't know which system returned each paper. Both measured the Consensus product rather than the APIs on the other comparison pages, so neither ranks PubMed, Europe PMC, OpenAlex or Semantic Scholar. To compare providers on your own domain, run a benchmark like [the one Phylo ran before choosing Consensus](/use-cases/benchmark-retrieval). ## Consensus vs Google Scholar Published 5 December 2025[1](#source-1). **What was tested.** Consensus Quick search, the single-pass ranked search, against Google Scholar. Consensus's agentic modes weren't part of this test. **Method.** * 500 real Consensus user queries, randomly sampled and anonymized, ranging from keyword lookups to natural-language questions. * The top 10 results from each system, about 10,000 query–paper pairs. * Three independent PhD researchers, with no affiliation to Consensus and no knowledge of which system returned each paper, scored every pair from 0 (irrelevant) to 4 (perfectly relevant) on its title and abstract. **Results.** | | Consensus | Google Scholar | | - | - | - | | Results scored relevant (3 or 4) | 88.1% | 81.8% | | Average DCG, which weights higher-ranked results more | 45.16 | 40.73 | Consensus also had higher average relevance at every rank from 1 to 10. ## Consensus vs GPT-5.5 and Opus 4.8 with web search Published 4 August 2026[2](#source-2). **What was tested.** Consensus Pro, the research agent that the [Threads API](/api-quickstart-threads) runs in `pro` mode, against GPT-5.5 and Opus 4.8, each with web search enabled and given the same query. **Method.** * 100 real Consensus user queries, randomly sampled and anonymized. * The top 20 results from each system. * Four independent PhD-level researchers, blind to the source, scored each pair from 0 to 4. LLM-as-a-judge across several frontier models gave a second, independent signal. * Precision is the share of results scored 3 or 4. Latency is wall-clock time from query to a complete set of results. **Results.** Consensus was more accurate than both frontier models, and far faster. Against GPT-5.5 the precision gain is modest, 1.8% in the top 10, but Consensus returned its results 8.6 times faster. Against Opus 4.8 it was both 11.6% more precise and 11.8 times faster. | | vs GPT-5.5 | vs Opus 4.8 | | - | - | - | | Median time to results | **8.7 s vs 74.8 s (8.6× faster)** | **8.7 s vs 102.8 s (11.8× faster)** | | Precision in the top 10 | 1.8% higher (1.4 points) | 11.6% higher (8.2 points) | | Precision in the top 20 | 4.5% higher (3.0 points) | 13.2% higher (8.2 points) | Even comparing Consensus's slower queries with the frontier models' faster ones, at the 75th percentile, Consensus was 3.8 and 5.1 times faster. The frontier models got there by brute force, running many more iterations and queries per result, so each answer also cost far more compute. Query by query, Consensus returned the most relevant results on 59 of the 100 queries, GPT-5.5 on 22 and Opus 4.8 on 6, with 11 ties. Consensus held relevance better deeper into the list: between rank 10 and rank 20 its precision fell 8.7 points, against 10.3 for GPT-5.5. ## What these evaluations don't show * **Relevance is textual.** Judges scored how well a paper matches the query, not its recency, citations or journal quality. * **Judging used title and abstract**, because some systems restrict full-text access. * **The frontier models had no budget.** They could spend as much time and as many tokens as they wanted on each query. Consensus returned more relevant results in a fraction of the time and compute. Time was measured directly; tokens and dollars weren't. ## Run your own benchmark The most convincing evaluation is one on your own queries. [Benchmark literature providers against your own ground truth](/use-cases/benchmark-retrieval) walks through it: curate queries across your domain, list the papers an expert would insist on, and measure how many each provider surfaces in its top 20. ## Related > [card] [How search works](/get-started/how-search-works) > > Hybrid search, ranking and reranking, step by step. > [card] [Compare scholarly APIs](/api-comparison) > > Why teams choose Consensus, and every API at a glance. ## Sources Facts about other providers were checked against their official documentation on 5 November 2026. Limits and prices change, so check the source before you build on a number. 1. [Consensus Outperforms Google Scholar for Academic Search Retrieval](https://consensus.app/home/blog/consensus-outperforms-google-scholar-for-academic-search-retrieval/), Consensus blog, 5 December 2025. 2. [Consensus Beats Claude and ChatGPT in Academic Search](https://consensus.app/home/blog/consensus-outperforms-frontier-models-in-scientific-search/), Consensus blog, 4 August 2026. # Open metadata APIs: Consensus, OpenAlex, Semantic Scholar and Crossref Source: https://docs.consensus.app/api-comparison/openalex-semantic-scholar-crossref How the Consensus API compares with OpenAlex, the Semantic Scholar API and the Crossref REST API: coverage, semantic search, evidence metadata, licences, and how to enrich Consensus results with open metadata. OpenAlex, Semantic Scholar and Crossref are open infrastructure for scholarly metadata across every discipline. They're the right tools for bibliometrics, citation graphs and identifiers. Consensus draws on OpenAlex and Semantic Scholar for its own metadata layer, then adds publisher full text, evidence classification and ranking built for finding the best answer to a research question. ![Pyramid showing three layers: abstracts and metadata across all publishers from Semantic Scholar, OpenAlex, and PubMed; all open-access full text; and full text of paywalled papers from partner publishers](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/papers-pyramid.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=c6dd63f78277c9e6af4af8144e9c65a9) The Consensus index: abstracts and metadata from Semantic Scholar, OpenAlex and PubMed for every paper, all open-access full text, and full text of paywalled papers from publisher partners. See [where the papers come from](/get-started/where-papers-come-from) for how each layer is built. ## What each API does best | | Best for | Coverage | | - | - | - | | **Consensus** | | | | **OpenAlex** | | | | **Semantic Scholar** | | | | **Crossref** | | | ## Semantic search compared OpenAlex and Semantic Scholar both offer some form of search by meaning, with limits worth knowing. * **OpenAlex semantic search** (`search.semantic`) embeds each work's title, abstract and venue name. It returns up to 50 results per query, at 1 request a second, and supports 17 filters. Filters such as `cited_by_count` aren't supported[7](#source-7). Its regular `search` matches title, abstract, full text and keywords[8](#source-8). * **Semantic Scholar** offers relevance search (up to 1,000 ranked results), Boolean bulk search, and snippet search, which returns passages of about 500 words from the paper body[9](#source-9). * **Crossref** searches bibliographic metadata[4](#source-4). * **Consensus** runs semantic and keyword search together over 400M+ articles and ranks for scientific relevance. Every filter combines with the search, and a query returns up to 1,000 results. ## Evidence metadata | | What each result tells you about the evidence | | - | - | | **Consensus** | | | **OpenAlex** | | | **Semantic Scholar** | | | **Crossref** | | ## Enrich Consensus results with open metadata Every Consensus result carries a `doi`. Use it to pull what the open APIs do best. DOI lookups are free in OpenAlex[11](#source-11). ```python Python theme={null} import os import requests papers = requests.get( "https://api.consensus.app/v1/search", headers={"x-api-key": os.environ["CONSENSUS_API_KEY"]}, params={"query": "does creatine improve working memory", "study_types": "rct,meta-analysis"}, timeout=30, ).json()["results"] for paper in papers: if not paper["doi"]: continue # Authors, institutions and funders from OpenAlex work = requests.get(f"https://api.openalex.org/works/doi:{paper['doi']}", timeout=30).json() # Citations from Semantic Scholar s2 = requests.get( f"https://api.semanticscholar.org/graph/v1/paper/DOI:{paper['doi']}", params={"fields": "citationCount,citations.title"}, timeout=30, ).json() print(paper["title"], paper["study_type"], work.get("cited_by_count"), s2.get("citationCount")) ``` ## Licences and terms * **OpenAlex:** all data is CC0[1](#source-1). * **Semantic Scholar:** data is licensed CC BY-NC or ODC-BY, and you must attribute Semantic Scholar. AI2 uses the data you send to the API to provide and improve it[12](#source-12). * **Crossref:** most metadata isn't subject to copyright and can be used for any purpose. Some abstracts are copyrighted by publishers or authors[4](#source-4). * **Consensus:** customer data isn't used to train AI models. See [Pricing, rate limits and data privacy](/api-comparison/pricing-and-privacy). ## The same question on each API > [tab] Consensus > > ```bash cURL theme={null} > curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ > --data-urlencode "query=does creatine improve working memory" \ > --data-urlencode "study_types=rct,meta-analysis" > ``` > [tab] OpenAlex > > ```bash cURL theme={null} > curl -G "https://api.openalex.org/works" \ > --data-urlencode "search=creatine working memory" \ > --data-urlencode "filter=study_designs.id:randomized-controlled-trial" \ > --data-urlencode "api_key=$OPENALEX_API_KEY" > ``` > > Semantic search (`search.semantic`) doesn't accept the `study_designs` filter[7](#source-7). > [tab] Semantic Scholar > > ```bash cURL theme={null} > curl -G "https://api.semanticscholar.org/graph/v1/paper/search" \ > --data-urlencode "query=creatine working memory" \ > --data-urlencode "publicationTypes=ClinicalTrial" \ > --data-urlencode "fields=title,year,publicationTypes,citationCount" \ > -H "x-api-key: $S2_API_KEY" > ``` > [tab] Crossref > > ```bash cURL theme={null} > curl -G "https://api.crossref.org/works" \ > --data-urlencode "query=creatine working memory" \ > --data-urlencode "mailto=you@example.com" > ``` > > Results include every work type, such as book chapters, so expect to filter. ## Related > [card] [Compare scholarly APIs](/api-comparison) > > Why teams choose Consensus, and every API at a glance. > [card] [Where the papers come from](/get-started/where-papers-come-from) > > How Consensus builds its index from open metadata, its own crawl and publisher full text. ## Sources Facts about other providers were checked against their official documentation on 5 November 2026. Limits and prices change, so check the source before you build on a number. 1. [OpenAlex pricing](https://help.openalex.org/access/pricing/): all data CC0; \$1 of free API usage a day with a key; annual plans from \$5,000 a year; the whole database can be downloaded as a snapshot refreshed every quarter. A keyless response notes that it shares a \$0.10 daily budget per IP address. 2. [OpenAlex works](https://help.openalex.org/data/works/): more than 320 million works in the core catalog. 3. [Semantic Scholar API overview](https://www.semanticscholar.org/product/api): 214 million papers; unauthenticated requests share 1,000 requests a second across all users and may be throttled under heavy use; the introductory limit for an API key is 1 request a second. The API also serves SPECTER2 embeddings and a Recommendations service for similar papers. 4. [Crossref REST API](https://www.crossref.org/documentation/retrieve-metadata/rest-api/): bibliographic metadata including funding, licence and abstracts; no sign-up required; most metadata isn't subject to copyright. 5. [Tips for using the Crossref REST API](https://www.crossref.org/documentation/retrieve-metadata/rest-api/tips-for-using-the-crossref-rest-api/): annual public data file. 6. `total-results` from `https://api.crossref.org/works?rows=0` on 5 November 2026: 188,499,975. 7. [OpenAlex semantic search](https://help.openalex.org/api/semantic-search/): embeds title, abstract and venue name; up to 50 results per query; 1 request a second; 17 supported filters, and others such as `cited_by_count` return a 400 error. 8. [OpenAlex search](https://help.openalex.org/api/searching/): `search=` matches works by title, abstract, full text and keywords; Boolean and phrase searches are supported. 9. [Semantic Scholar Academic Graph API reference](https://api.semanticscholar.org/api-docs/graph): `/paper/search` (up to 1,000 relevance-ranked results), `/paper/search/bulk` (Boolean), `/snippet/search` (passages of about 500 words from title, abstract and body), and the `publicationTypes`, `minCitationCount` and `venue` filters. 10. [OpenAlex study designs](https://help.openalex.org/data/study-designs/): seven designs tagged automatically from title and abstract; each value right at least 96% of the time. 11. [OpenAlex single entities](https://help.openalex.org/api/get-single-entities/): lookups by ID or DOI, such as `/works/doi:10.7717/peerj.4375`, are free. 12. [Semantic Scholar API License Agreement](https://www.semanticscholar.org/product/api/license): AI2 uses the data you send to call the API "for the purpose of providing and improving the API"; data is licensed CC BY-NC or ODC-BY; attribution required. # Pricing, rate limits and data privacy Source: https://docs.consensus.app/api-comparison/pricing-and-privacy Pricing, rate limits, interfaces and data use compared across the Consensus API, PubMed E-utilities, Europe PMC, OpenAlex, Semantic Scholar and Crossref, including what each provider does with your queries. ## Pricing | | Free usage | Paid options | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | | **OpenAlex** | | | | **Semantic Scholar** | | | | **Crossref** | | | A dash (—) means we didn't find a paid option in that provider's documentation as of November 2026. ### How Consensus counts calls * One call covers up to 100 papers returned, rounded up, with a minimum of one call per request. A request for 20 papers is 1 call. A request for 150 papers is 2 calls. * The API and the MCP server share one monthly pool. * Usage resets on the 1st of each month. See [plans and access](/api-plans-and-access) for page sizes and features by plan. ## Rate limits | | Without a key | With a key or plan | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | | **OpenAlex** | | | | **Semantic Scholar** | | | | **Crossref** | | | ## Interfaces and code samples | | Interfaces | | - | - | | **Consensus** | | | **PubMed** | | | **Europe PMC** | | | **OpenAlex** | | | **Semantic Scholar** | | | **Crossref** | | ## Data privacy: what happens to your queries What you search for can reveal your pipeline: drug targets, materials, competitor products. Check what a provider does with your queries before you send them. ### Consensus * **SOC 2 Type II** compliant, hosted on Google Cloud. * **Customer data isn't used to train AI models.** The LLM providers Consensus uses keep no data from requests. * **Zero data retention** and a **data-processing agreement** are available on Enterprise contracts. * API keys are shown once, and can be named, revoked and rotated from the dashboard. See [Security and data handling](/enterprise/security) for encryption, infrastructure and access controls. ### Other providers The free APIs publish data licences and usage terms rather than enterprise privacy commitments: * **PubMed** asks every caller to identify themselves with `tool` and `email` parameters[8](#source-8). * **Europe PMC** serves open-access content and metadata, and doesn't allow automated bulk download of other content[11](#source-11). * **OpenAlex** releases all its data under CC0. API keys are tied to an account[4](#source-4). * **Semantic Scholar**'s licence says AI2 uses the data you send to call the API "for the purpose of providing and improving the API". Its data is licensed CC BY-NC or ODC-BY, and attribution is required[14](#source-14). * **Crossref** asks for an email in the polite pool. Most of its metadata isn't subject to copyright[6](#source-6). Read each provider's terms against your own policy. ## Related > [card] [API plans and access](/api-plans-and-access) > > Calls, page sizes, features and errors by plan. > [card] [Compare scholarly APIs](/api-comparison) > > Why teams choose Consensus, and every API at a glance. ## Sources Facts about other providers were checked against their official documentation on 5 November 2026. Limits and prices change, so check the source before you build on a number. 1. [About PubMed](https://pubmed.ncbi.nlm.nih.gov/about/): "more than 40 million citations and abstracts of biomedical literature"; "does not include full text journal articles"; a free resource. 2. [About Europe PMC](https://europepmc.org/About): 49.2 million abstracts and 12.5 million full-text articles; searches abstracts and full text; searches can be limited to clinical trials, preprints or article sections such as Methods; free to anyone. 3. [OpenAlex example costs](https://help.openalex.org/access/example-costs/): search costs \$1 per 1,000 calls; content downloads cost \$10 per 1,000. 4. [OpenAlex pricing](https://help.openalex.org/access/pricing/): all data CC0; \$1 of free API usage a day with a key; annual plans from \$5,000 a year; the whole database can be downloaded as a snapshot refreshed every quarter. A keyless response notes that it shares a \$0.10 daily budget per IP address. 5. [Semantic Scholar API overview](https://www.semanticscholar.org/product/api): 214 million papers; unauthenticated requests share 1,000 requests a second across all users and may be throttled under heavy use; the introductory limit for an API key is 1 request a second. The API also serves SPECTER2 embeddings and a Recommendations service for similar papers. 6. [Crossref REST API](https://www.crossref.org/documentation/retrieve-metadata/rest-api/): bibliographic metadata including funding, licence and abstracts; no sign-up required; most metadata isn't subject to copyright. 7. [Crossref access and authentication](https://www.crossref.org/documentation/retrieve-metadata/rest-api/access-and-authentication/): public pool 5 requests a second, polite pool 10, Metadata Plus 150. 8. [A General Introduction to the E-utilities](https://www.ncbi.nlm.nih.gov/books/NBK25497/): more than 3 requests a second without an API key returns an error; with a key, up to 10 a second by default; higher rates by request. Requests without registered `tool` and `email` values may be blocked. 9. Europe PMC team on the [Europe PMC web services forum](https://groups.google.com/a/ebi.ac.uk/g/epmc-webservices/c/x3PLL09mels), 1 October 2024: the rate limit is 10 requests a second (500 a minute), and no API key is required. 10. [OpenAlex authentication and rate limits](https://help.openalex.org/api/authentication/): 10 requests a second with no key, 30 with a free key or prepaid usage, 50 on an annual plan; semantic search is limited to 1 a second. 11. [Europe PMC developer resources](https://europepmc.org/developers): REST, SOAP, OAI-PMH and FTP access to open-access content and metadata; automated bulk download of other content isn't allowed. The [Articles RESTful API](https://europepmc.org/RestfulWebService) includes `/{id}/fullTextXML`. 12. [OpenAlex help center index](https://help.openalex.org/llms.txt): REST URLs, OQL, a CLI, and Claude and ChatGPT connectors. 13. [Tips for using the Crossref REST API](https://www.crossref.org/documentation/retrieve-metadata/rest-api/tips-for-using-the-crossref-rest-api/): annual public data file. 14. [Semantic Scholar API License Agreement](https://www.semanticscholar.org/product/api/license): AI2 uses the data you send to call the API "for the purpose of providing and improving the API"; data is licensed CC BY-NC or ODC-BY; attribution required. # Biomedical search: Consensus, PubMed and Europe PMC Source: https://docs.consensus.app/api-comparison/pubmed-europe-pmc How the Consensus API compares with PubMed E-utilities and the Europe PMC REST API for biomedical literature search: semantic search, study-design filters, clinical guidelines, full text, and when a Boolean strategy is the right tool. PubMed and Europe PMC are the reference sources for biomedical literature. PubMed, run by the US National Library of Medicine, holds more than 40 million citations and abstracts[1](#source-1). Europe PMC, run by EMBL-EBI, holds 49.2 million abstracts and 12.5 million full-text articles, including preprints[2](#source-2). Both are free, and both are excellent for exhaustive Boolean searching. Consensus covers all of PubMed and the rest of science, and is built for a different job: finding the strongest evidence for a question asked in plain language. It searches full text from top medical publishers such as BMJ, JAMA Network and Wolters Kluwer, and its research agent has tools that find better papers than a single query can: it crawls citations from key studies, traces an author's work and plans multi-step searches. ![Pyramid showing three layers: abstracts and metadata across all publishers from Semantic Scholar, OpenAlex, and PubMed; all open-access full text; and full text of paywalled papers from partner publishers](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/papers-pyramid.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=c6dd63f78277c9e6af4af8144e9c65a9) All of PubMed is one source in the Consensus index, alongside Semantic Scholar, OpenAlex, all open-access full text, and paywalled full text from publisher partners. ## What each API does best | | Best for | Watch out for | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | ## Semantic search over PubMed's records PubMed matches the terms in your query, mapping them to MeSH headings where it can, and can sort the matches with its Best Match algorithm[3](#source-3), [5](#source-5). That works well when you know the vocabulary. A question worded differently from the abstracts that answer it can still miss them. Consensus searches the same records by meaning and by keyword together, then ranks the results for scientific relevance. You send the question as written, for example "does creatine improve working memory in older adults", and get the most relevant papers first. Boolean operators and exact phrases still work when you want them. Europe PMC also matches keywords, but against full text as well as abstracts, so it can find papers whose abstracts never use your terms[2](#source-2). ## Filtering by study design and evidence quality All three APIs let you narrow results by study type. Consensus adds the other signals a reviewer looks at. | | Study design | Other evidence filters | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | Each Consensus result also returns its `study_type`, `sample_size`, `population_type` and `sjr_best_quartile`, so your code can sort or grade the evidence after the search. ## Top medical journals and clinical guidelines Two Consensus filters are built for clinical work: * **`medical_mode=true`** limits results to about 8 million documents from the top 1,000 medical journals, plus about 50,000 clinical guidelines. Use it when general literature is too noisy. * **`clinical_guideline=true`** returns only papers classified as clinical guidelines. ## When a Boolean strategy is the right tool For the search you report in a systematic review, use a Boolean query in PubMed or Europe PMC. Reviewers can rerun it term for term, and it returns every match rather than a ranked top 1,000. Consensus fits around that search: scope the topic before you write the strategy, then check afterwards that the strongest studies on the question all made it into your set. ## Full text * **Consensus** searches the full text of open-access papers and of paywalled papers from its publisher partners. On paid plans, each search also returns up to 3 query-relevant passages per paper from open-access papers. See [Full text in the API](/api-full-text). * **PubMed** has no full text. Its sister archive, PubMed Central, holds more than 10 million full-text records[7](#source-7). * **Europe PMC** returns open-access articles as XML through `fullTextXML`[4](#source-4). ## The same question on each API Consensus takes the question as written. PubMed and Europe PMC need it rewritten as terms and filters. > [tab] Consensus > > ```bash cURL theme={null} > curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ > --data-urlencode "query=does creatine improve working memory" \ > --data-urlencode "study_types=rct,meta-analysis" \ > --data-urlencode "human=true" > ``` > > ```python Python theme={null} > import os > import requests > > resp = requests.get( > "https://api.consensus.app/v1/search", > headers={"x-api-key": os.environ["CONSENSUS_API_KEY"]}, > params={ > "query": "does creatine improve working memory", > "study_types": "rct,meta-analysis", > "human": True, > }, > timeout=30, > ) > resp.raise_for_status() > > for paper in resp.json()["results"]: > print(paper["study_type"], paper["sample_size"], paper["title"]) > ``` > [tab] PubMed > > ```bash cURL theme={null} > curl -G "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi" \ > --data-urlencode "db=pubmed" \ > --data-urlencode "term=creatine AND memory AND randomized controlled trial[pt]" \ > --data-urlencode "sort=relevance" \ > --data-urlencode "retmode=json" > ``` > > Returns PubMed IDs. Fetch titles and abstracts with a second call to `esummary` or `efetch`. > [tab] Europe PMC > > ```bash cURL theme={null} > curl -G "https://www.ebi.ac.uk/europepmc/webservices/rest/search" \ > --data-urlencode 'query=creatine AND memory AND PUB_TYPE:"randomized controlled trial"' \ > --data-urlencode "format=json" > ``` ## Related > [card] [Compare scholarly APIs](/api-comparison) > > Why teams choose Consensus, and every API at a glance. > [card] [Swap a hand-maintained paper corpus](/use-cases/swap-a-paper-corpus) > > How Owkin replaced a curated set of PubMed abstracts with Consensus. ## Sources Facts about other providers were checked against their official documentation on 5 November 2026. Limits and prices change, so check the source before you build on a number. 1. [About PubMed](https://pubmed.ncbi.nlm.nih.gov/about/): "more than 40 million citations and abstracts of biomedical literature"; "does not include full text journal articles"; a free resource. 2. [About Europe PMC](https://europepmc.org/About): 49.2 million abstracts and 12.5 million full-text articles; searches abstracts and full text; searches can be limited to clinical trials, preprints or article sections such as Methods; free to anyone. 3. [PubMed User Guide](https://pubmed.ncbi.nlm.nih.gov/help/): Boolean operators, field tags, MeSH and Best Match sorting. 4. [Europe PMC developer resources](https://europepmc.org/developers): REST, SOAP, OAI-PMH and FTP access to open-access content and metadata; automated bulk download of other content isn't allowed. The [Articles RESTful API](https://europepmc.org/RestfulWebService) includes `/{id}/fullTextXML`. 5. [E-utilities in depth, ESearch](https://www.ncbi.nlm.nih.gov/books/NBK25499/): `sort=relevance` is the default Best Match order on web PubMed. 6. [NLM publication types](https://www.nlm.nih.gov/mesh/pubtypes.html). 7. [About PMC](https://pmc.ncbi.nlm.nih.gov/about/intro/): "more than 10 million full-text article records". # Retrieval for RAG and AI agents Source: https://docs.consensus.app/api-comparison/rag-and-ai-agents What a RAG pipeline or research agent needs from a scholarly search API, how Consensus compares with PubMed, Europe PMC, OpenAlex and Semantic Scholar for grounding, and when web search APIs like Exa and Tavily fit. A RAG pipeline or research agent is only as good as what its retrieval step puts in the context window. For scientific questions, that means four things from a search API: 1. **The right papers in the top few results** for a question written in plain language, because agents ask in sentences. 2. **Full-text snippets, not just titles**, so the model reads the passage that supports a claim. 3. **Metadata to weigh the evidence**, so the model can prefer a meta-analysis to a case report. 4. **Stable identifiers**, so every claim the model makes stays citable. ## How the APIs compare for RAG | | Plain-language queries | Text for the context window | | - | - | - | | **Consensus** | | | | **PubMed** | | | | **Europe PMC** | | | | **OpenAlex** | | | | **Semantic Scholar** | | | ## Full-text snippets, ready for a prompt On a paid plan, every Consensus search returns these snippets by default. Each eligible result gains up to 3 passages from the body of the paper, chosen for your query, each starting with the section it came from. You get the doses, methods and measured values that abstracts leave out, without downloading and chunking PDFs. ```python Python theme={null} import os import requests resp = requests.get( "https://api.consensus.app/v1/search", headers={"x-api-key": os.environ["CONSENSUS_API_KEY"]}, params={"query": "what dose of creatine improves working memory"}, timeout=30, ) resp.raise_for_status() context = [] for paper in resp.json()["results"]: evidence = paper["full_text_chunks"] or [paper["abstract"]] for passage in evidence: context.append(f"[{paper['doi']}] ({paper['study_type']}, n={paper['sample_size']}) {passage}") ``` Snippets come from open-access papers. Partner full text improves which papers rank, but isn't returned. When a paper has no snippet, fall back to its `abstract` and `takeaway`. See [Full text in the API](/api-full-text). ## Or let Consensus run the agent If you'd rather not build the retrieval loop, the [Threads API](/api-quickstart-threads) runs the Consensus research agent for you. It does more than search: it plans a multi-step strategy, crawls citations backward and forward from key papers, looks up authors, finds specific papers and reads them in full. It returns a written answer with numbered citations and the quotes it relied on. * **`pro`** reads about 20 papers. Use it for most questions. * **`deep`** reads about 50 papers and writes a structured report. It needs a paid plan. For assistants and agents that speak MCP, the [Consensus MCP server](/consensus-mcp) offers the same search and research agent as tools. ## Web search APIs vs a scholarly search API Exa and Tavily are web search APIs built for AI agents. Exa's search endpoint "lets you search the web", and offers a `publication` category for research papers and journal articles[8](#source-8). Tavily calls itself "the web layer for AI agents"[9](#source-9). They're the right tools for news, company pages and the open web. The biggest difference is full text. Consensus searches the full text of papers from 13 publisher partners, including 7 of the 10 largest academic publishers, as well as every open-access paper, and hands your agent the passage from the paper that supports the point. That text isn't on the open web, so a web index can't search it. Search results carry snippets from open-access papers, and the research agent's citations carry the quotes it relied on. A scholarly API also gives your agent: * **A defined corpus of papers**, each with a DOI, journal and year, instead of a mix of papers, press releases and blog posts. * **Evidence metadata**, so the agent can tell a randomized trial from a case report and a peer-reviewed paper from a preprint. * **Retraction handling.** Consensus marks retracted papers and never uses them in its AI answers. Many teams give their agent both: web search for current events and general facts, Consensus for anything that needs a citation to the literature. ## In production * **Owkin** replaced a hand-maintained set of PubMed abstracts with the Consensus MCP server inside its AI scientist, and got 25x the paper coverage. See [Swap a hand-maintained paper corpus](/use-cases/swap-a-paper-corpus). * **Phylo**'s agents fan each biology question out into parallel Consensus API searches to screen about 20,000 drug targets. See [Parallel search angles](/use-cases/parallel-search-angles). ## Related > [card] [Add an evidence layer to an agent](/use-cases/add-an-evidence-layer) > > Give a working agent peer-reviewed literature as a tool. > [card] [Compare scholarly APIs](/api-comparison) > > Why teams choose Consensus, and every API at a glance. ## Sources Facts about other providers were checked against their official documentation on 5 November 2026. Limits and prices change, so check the source before you build on a number. 1. [PubMed User Guide](https://pubmed.ncbi.nlm.nih.gov/help/): Boolean operators, field tags, MeSH and Best Match sorting. 2. [About PMC](https://pmc.ncbi.nlm.nih.gov/about/intro/): "more than 10 million full-text article records". 3. [About Europe PMC](https://europepmc.org/About): 49.2 million abstracts and 12.5 million full-text articles; searches abstracts and full text; searches can be limited to clinical trials, preprints or article sections such as Methods; free to anyone. 4. [Europe PMC developer resources](https://europepmc.org/developers): REST, SOAP, OAI-PMH and FTP access to open-access content and metadata; automated bulk download of other content isn't allowed. The [Articles RESTful API](https://europepmc.org/RestfulWebService) includes `/{id}/fullTextXML`. 5. [OpenAlex semantic search](https://help.openalex.org/api/semantic-search/): embeds title, abstract and venue name; up to 50 results per query; 1 request a second; 17 supported filters, and others such as `cited_by_count` return a 400 error. 6. [OpenAlex full text](https://help.openalex.org/access/fulltext/): cached PDF and TEI XML for more than 60 million open-access works. 7. [Semantic Scholar Academic Graph API reference](https://api.semanticscholar.org/api-docs/graph): `/paper/search` (up to 1,000 relevance-ranked results), `/paper/search/bulk` (Boolean), `/snippet/search` (passages of about 500 words from title, abstract and body), and the `publicationTypes`, `minCitationCount` and `venue` filters. 8. [Exa search API reference](https://docs.exa.ai/reference/search): "The search endpoint lets you search the web"; the `publication` category covers research papers, preprints and journal articles. 9. [About Tavily](https://docs.tavily.com/documentation/about): "the web layer for AI agents". # Full text in the API Source: https://docs.consensus.app/api-full-text Read inside papers with the Consensus API: query-matched excerpts from open-access papers in every paid search, and the complete text of open-access papers as Markdown. > [!TIP] > > New to the API? Make your first request with the [quickstart](/api-get-started#quickstart), then come back here. Abstracts leave out most of what you need: doses, methods, measured values, limitations. The Consensus API returns the text inside open-access papers in two ways: excerpts that come back with every search, and the complete text of the papers you pick. | | Excerpts | Complete text | | - | - | - | | **What you get** | Up to 3 passages per paper that match your query, labeled by section | The whole paper as Markdown, with a list of its sections | | **How to get it** | Returned with every paid search: read `full_text_chunks` on each search result | `GET /v1/papers/full_text` with 1 to 20 paper ids or DOIs. In MCP, `get_full_text` | | **Which papers** | Open access only. [Threads](#which-papers-have-excerpts) can also quote paywalled papers they cite | Open access only. Other papers come back `blocked` | | **Plans** | Paid plans (Pro, Deep, Teams, Enterprise) | Paid plans (Pro, Deep, Teams, Enterprise) | | **Cost** | Included in the search | 1 credit per paper returned | Search itself also reads full text, on every plan: queries are matched against titles, abstracts, and full text where Consensus holds it, so a paper can rank on what's in its methods or results. See [how search works](/get-started/how-search-works). ## Excerpts ### Example On a paid plan, open-access results carry excerpts in `full_text_chunks`: ```bash cURL theme={null} curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=what dose of creatine improves working memory" ``` ```python Python theme={null} import os, requests resp = requests.get( "https://api.consensus.app/v1/search", headers={"x-api-key": os.environ["CONSENSUS_API_KEY"]}, params={"query": "what dose of creatine improves working memory"}, timeout=30, ) resp.raise_for_status() for paper in resp.json()["results"]: for chunk in paper.get("full_text_chunks") or []: section, _, passage = chunk.partition(" | ") print(paper["doi"], "·", section.removeprefix("Section: "), "·", passage[:120]) ``` A real response, trimmed to one result with shortened excerpts: ```json wrap theme={null} { "results": [ { "title": "Creatine and improvement in cognitive function: Evaluation of a health claim pursuant to article 13(5) of regulation (EC) No 1924/2006", "journal_name": "EFSA Journal", "publish_year": 2024, "doi": "10.2903/j.efsa.2024.9100", "url": "https://consensus.app/papers/creatine-and-improvement-in-cognitive-function-turck-bohn/7e703e828b815257b31687a25eceac9d/", "takeaway": "Creatine supplementation has not been established as a cause-and-effect relationship for improving cognitive function in one or more domains.", "full_text_chunks": [ "Section: Human intervention studies in healthy individuals conducted under normal conditions | Two studies had an initial loading phase of 20 g/day of creatine for 5 days, followed by 5 g/day … A significant improvement in working memory, as measured by the Backwards Digit Span test, was observed only in the study of shorter duration …", "Section: Weighing of the evidence | … the acute effect of creatine on working memory reported in some studies at doses of 20 g/day given for 5-7 days … was not observed at lower creatine doses, ranging from 2.2 to 14 g/day …" ] } ] } ``` The `takeaway` gives you the paper's conclusion. The excerpts give you the doses and study designs behind it, which the abstract doesn't. > [!NOTE] > > As of November 2026, you no longer need to send `include_full_text_chunks=true`: paid plans return excerpts by default. Send `include_full_text_chunks=false` to leave them out when you only need metadata. ### What comes back > [accordion] Format > > `full_text_chunks` is a list of strings. Each one starts with the section of the paper it came from, then the passage: `Section: | `. Split on the first `|` to separate them. > [accordion] How many excerpts > > Up to **3 excerpts per paper**, for the first **20 eligible papers** on each page, so at most 60 per request. Later papers on a larger page carry none. Excerpts are the passages most relevant to your query across those papers, so some may carry fewer than 3, or none. To get excerpts across more papers, keep `page_size` at 20 and page through with `page`. > [accordion] When a paper has none > > A result has no `full_text_chunks` field when the paper doesn't meet the [open-access rule](#which-papers-have-excerpts), Consensus has no indexed text for it, no passage matched your query, or it comes after the first 20 eligible papers on the page. Fall back to `abstract` and `takeaway` for those papers. Excerpts never fail a search: if the excerpt lookup errors or takes too long, the results come back without them. ### Which papers have excerpts Excerpts come only from **open-access** papers, whatever your plan. A paper qualifies when both of these hold: * **Its open-access status is gold, diamond, green, or hybrid:** the publisher's version of the article, or a repository copy, is open. * **The text Consensus holds came from an open source:** arXiv, CSHL (bioRxiv, medRxiv), OpenAlex, PubMed Central, or S2ORC. Text from a publisher feed never qualifies, whatever the paper's open-access status. Bronze papers (free to read on the publisher's site, but without an open license), closed papers, and papers whose status is unknown never carry excerpts. Neither do retracted papers. > [!NOTE] > > Consensus also reads paywalled full text from its [publisher partners](/get-started/full-text-access). Search excerpts and complete text never return it, even when your organization has access to those publishers. [Threads](/api-quickstart-threads) are the exception: when a Thread's analysis cites a paywalled paper, that citation's grounding quotes (`citations[].snippets`) can come from the paper's full text. When excerpts matter more than coverage, add `open_access=true` to limit results to open-access papers, so more of each page can carry them. ### Excerpts by plan Excerpts are available on all paid plans: Pro, Deep, Teams, and Enterprise. Retrieval over full text works on every plan. Free plan searches return no excerpts, and `include_full_text_chunks=true` on a Free plan returns `403` with the code `feature_not_allowed`. The [MCP server](/consensus-mcp)'s [`search`](/mcp-tools/search) tool returns excerpts on paid plans too, up to 3 per paper for the first 5 eligible papers, and 12,000 characters per search, so an assistant can quote the passage behind each claim. ### Get the most out of excerpts * **Phrase the query as the thing you want to find.** Excerpts are chosen against your query, so "what dose of creatine improves working memory" returns dosing passages where "creatine" returns general ones. * **Keep each excerpt with its paper.** Store the `doi` and `url` alongside every passage so an answer built on it stays citable. * **Say which evidence you used.** When a result has no `full_text_chunks` and you fall back to the abstract, tell the reader. * **Read the paper for anything high-stakes.** Excerpts are passages, not whole papers. Request its [complete text](#complete-text) when it's open access, or follow `url` to the paper. ## Complete text When a passage isn't enough, read the whole paper: every method detail, every results table, the full discussion. Search first, pick the papers you want, then request their complete text. **1. Search** Every search result carries a `paper_id` and `full_text_available`. `full_text_available` is `true` when the paper is open access. It doesn't promise text: a paper that was retracted, or whose copy in Consensus isn't its open-access copy, comes back `blocked`, and a paper Consensus holds no text for comes back `no_full_text`. **2. Pick papers** Collect the `paper_id` of each paper you want to read, up to 20 per request. You can also name papers by DOI, or pass the `citations[].paper_id` values from a [Thread](/api-quickstart-threads). **3. Request the text** Call `GET /v1/papers/full_text` with one `paper_ids` or `dois` parameter per paper. Each paper whose status is `available` comes back as Markdown, with a list of its sections. ```bash cURL theme={null} API=https://api.consensus.app/v1 # 1. Search. Each result carries paper_id and full_text_available. curl -G "$API/search" -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=what dose of creatine improves working memory" # 2. Request the complete text of the papers you picked, by id or DOI. curl -G "$API/papers/full_text" -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "paper_ids=PAPER_ID_1" \ --data-urlencode "paper_ids=PAPER_ID_2" \ --data-urlencode "dois=10.2903/j.efsa.2024.9100" ``` ```python Python theme={null} import os import requests API = "https://api.consensus.app/v1" HEADERS = {"x-api-key": os.environ["CONSENSUS_API_KEY"]} # 1. Search, and keep the papers whose complete text can be returned. search = requests.get( f"{API}/search", headers=HEADERS, params={"query": "what dose of creatine improves working memory"}, timeout=30, ) search.raise_for_status() paper_ids = [p["paper_id"] for p in search.json()["results"] if p["full_text_available"]][:5] if not paper_ids: raise SystemExit("No open-access papers in these results.") # 2. Request their complete text. requests repeats paper_ids once per list item. resp = requests.get( f"{API}/papers/full_text", headers=HEADERS, params={"paper_ids": paper_ids}, timeout=60, ) resp.raise_for_status() body = resp.json() for paper in body["results"]: if paper["status"] == "available": titles = [s["title"] for s in paper["sections"]] print(paper["paper_id"], paper["total_chars"], "chars:", ", ".join(titles)) print(paper["markdown"][:300]) else: print(paper.get("paper_id") or paper["doi"], paper["status"]) print("Credits charged:", body["credits_charged"]) ``` ```javascript JavaScript theme={null} const API = "https://api.consensus.app/v1"; const headers = { "x-api-key": process.env.CONSENSUS_API_KEY }; // 1. Search, and keep the papers whose complete text can be returned. const query = new URLSearchParams({ query: "what dose of creatine improves working memory" }); const search = await fetch(`${API}/search?${query}`, { headers }); const { results } = await search.json(); const paperIds = results .filter((paper) => paper.full_text_available) .map((paper) => paper.paper_id) .slice(0, 5); if (paperIds.length === 0) throw new Error("No open-access papers in these results."); // 2. Request their complete text, with one paper_ids parameter per paper. const params = new URLSearchParams(); for (const id of paperIds) params.append("paper_ids", id); const resp = await fetch(`${API}/papers/full_text?${params}`, { headers }); if (!resp.ok) throw new Error(`${resp.status}: ${await resp.text()}`); const body = await resp.json(); for (const paper of body.results) { if (paper.status === "available") { const titles = paper.sections.map((section) => section.title); console.log(paper.paper_id, paper.total_chars, "chars:", titles.join(", ")); console.log(paper.markdown.slice(0, 300)); } else { console.log(paper.paper_id ?? paper.doi, paper.status); } } console.log("Credits charged:", body.credits_charged); ``` The response has one result per paper you named, paper ids first and then DOIs, each in the order you sent them: ```json wrap theme={null} { "results": [ { "paper_id": "", "status": "available", "free_reread_until": "2026-10-29T18:00:00Z", "open_access_status": "gold", "markdown": "## Introduction\n\nCreatine is …\n\n## Methods\n\n…", "sections": [ { "title": "Introduction", "offset": 0 }, { "title": "Methods", "offset": 4210 }, { "title": "Results", "offset": 12877 }, { "title": "Discussion", "offset": 27105 } ], "total_chars": 38412 }, { "paper_id": "", "status": "blocked", "open_access_status": "closed" }, { "paper_id": "7e703e828b815257b31687a25eceac9d", "doi": "10.2903/j.efsa.2024.9100", "status": "available", "free_reread_until": "2026-10-29T18:00:00Z", "open_access_status": "gold", "markdown": "…\n\n## Weighing of the evidence\n\n…", "sections": [ { "title": "Human intervention studies in healthy individuals conducted under normal conditions", "offset": 9318 }, { "title": "Weighing of the evidence", "offset": 21544 } ], "total_chars": 30187 } ], "credits_charged": 2 } ``` Illustrative and trimmed. Two open-access papers came back, so the request cost 2 credits. The blocked paper cost nothing. ### What comes back > [accordion] Markdown and sections > > `markdown` is the paper's complete text, with a `## ` heading wherever a new section starts. `sections` lists each heading with its `offset`, the character position where its heading line starts in `markdown`, so you can slice out one section. `total_chars` is the length of `markdown`. All three come back on every `available` paper. > [accordion] DOIs > > Send DOIs with or without a `https://doi.org/` or `doi:` prefix. A DOI result echoes the DOI, lowercased and without its prefix, plus the `paper_id` it resolved to. You can mix `paper_ids` and `dois` in one request, up to 20 combined. Naming the same paper twice returns it twice and charges it once. > [accordion] Errors > > `422` when a request names no papers, more than 20, or a value that isn't a paper id from search or Thread results. `403` `feature_not_allowed` on a Free plan. `429` when you send more than one request per second, or when your remaining monthly allowance can't cover every paper in the request. `503` when complete text is busy or can't look up or read papers right now; when it's busy, the response carries `Retry-After: 1`. Retry a `503` after a short wait. Nothing is charged for a refused request. ### Which papers have complete text Complete text follows the same open-access rule as [excerpts](#which-papers-have-excerpts): gold, diamond, green, and hybrid papers whose text came from an open source such as arXiv or PubMed Central, never from a publisher feed. Your plan and your organization's publisher access don't change which papers qualify. Each result's `status` says what happened to it: | `status` | Meaning | Charged | | - | - | - | | `available` | The paper is open access and Consensus holds its text. Its complete text comes back as Markdown. | 1 credit | | `blocked` | The paper isn't open access, the copy Consensus holds isn't its open-access copy, or it was retracted, so no text comes back. `open_access_status` shows the paper's status when it's known, so it can still be `gold` or `green`. | No | | `no_full_text` | The paper is open access, but Consensus holds no text for it. | No | | `not_found` | No Consensus paper has this id or DOI. | No | Check `full_text_available` on search results before you request, so you skip papers that are certain to come back blocked. ### Credits Complete text costs **1 credit per paper returned**. A credit counts the same as a call against your plan's monthly allowance, which the API and MCP share. * **Free:** papers that come back blocked, not found, or `no_full_text`. * **Charged once a day:** a paper charged to you in the last 24 hours, through the API or MCP, comes back again for free, so retrying after a response doesn't charge twice. This is best effort: a reread before `free_reread_until` can still be charged, for example when the first charge didn't complete. A request for the same paper sent before the first one has finished is charged too, so wait for a response before you retry. * **Hold first, settle after:** before reading, the request reserves 1 credit for each paper that could be charged. If that doesn't fit in your remaining allowance, the whole request returns `429` and nothing is charged. After the response, you pay only for the papers returned. `credits_charged` shows what the request is priced at; the usage it counts can be lower, since a request that crosses your included allowance is counted only up to it, and a charge that fails counts nothing. Complete text needs a paid plan: Pro, Deep, Teams, or Enterprise. On a Free plan the request returns `403` `feature_not_allowed`. See [plans and access](/api-plans-and-access#how-calls-are-counted). ### In MCP: search, then get_full_text The [MCP server](/consensus-mcp) offers the same complete text through the [`get_full_text`](/mcp-tools/get-full-text) tool, for every MCP client. Each paper in a `search` result shows an `id:` line, marked `full text available` when the paper is open access and `get_full_text` is available to you. Your assistant passes up to 3 of those ids, or DOIs, to `get_full_text` and gets each paper back as Markdown, up to 40,000 characters at a time. A longer paper ends with a cursor that reads the next part, free within the 24-hour re-read window. It costs the same 1 credit per paper returned. ## Related > [card] [Extract numbers from full-text papers](/use-cases/extract-values-from-full-text) > > A worked example that pulls values and their conditions out of excerpts. > [card] [Full text access](/get-started/full-text-access) > > Where Consensus's full text comes from, and how the app uses it. > [card] [Full text endpoint reference](/api-reference/get-full-text) > > Every parameter and response field of `GET /v1/papers/full_text`. > [card] [Search endpoint reference](/api-reference/search) > > Every search parameter, including `paper_id` and `full_text_available` on results. # Get started with Consensus API Source: https://docs.consensus.app/api-get-started Search over 400 million scholarly articles, including 350 million peer-reviewed research papers, with the Consensus REST API. The Consensus API lets you search 400+ million scholarly articles, including 350 million peer-reviewed research papers, over a single REST endpoint. Send a query and get back ranked papers with titles, authors, journals, publication years, citation counts, and direct links. ## Quickstart Make your first API call in a few minutes. Every plan includes API access, and the Free plan includes 30 calls a month. **1. Sign in to Consensus** Create a free account or sign in at [consensus.app](https://consensus.app). The dashboard in the next step only opens when you're signed in. **2. Open the API & MCP Dashboard** Go to [consensus.app/api-mcp](https://consensus.app/api-mcp/). You can also get there from your profile icon: choose **API & MCP Dashboard**. **3. Create and copy an API key** Click **Create API key**, name the key, and click **Copy API key**. The full key is shown only once, so store it somewhere safe. Already have a key? Open **Keys and clients** and click **New key**. **4. Make your first request** Save the key in an environment variable named `CONSENSUS_API_KEY`: ```bash theme={null} export CONSENSUS_API_KEY="paste-your-key-here" ``` Then search. `query` is the only required parameter, and the key goes in the `x-api-key` header. ```bash cURL theme={null} curl "https://api.consensus.app/v1/search?query=creatine%20and%20cognitive%20function" \ -H "x-api-key: $CONSENSUS_API_KEY" ``` ```python Python theme={null} import os import requests resp = requests.get( "https://api.consensus.app/v1/search", headers={"x-api-key": os.environ["CONSENSUS_API_KEY"]}, params={"query": "creatine and cognitive function"}, ) resp.raise_for_status() for paper in resp.json()["results"]: print(paper["title"], "—", paper["journal_name"], paper["publish_year"]) ``` ```javascript JavaScript theme={null} const params = new URLSearchParams({ query: "creatine and cognitive function" }); const resp = await fetch(`https://api.consensus.app/v1/search?${params}`, { headers: { "x-api-key": process.env.CONSENSUS_API_KEY }, }); const { results } = await resp.json(); for (const paper of results) { console.log(paper.title, "—", paper.journal_name, paper.publish_year); } ``` ```go Go theme={null} package main import ( "encoding/json" "fmt" "net/http" "net/url" "os" ) func main() { req, _ := http.NewRequest("GET", "https://api.consensus.app/v1/search", nil) req.Header.Set("x-api-key", os.Getenv("CONSENSUS_API_KEY")) req.URL.RawQuery = url.Values{"query": {"creatine and cognitive function"}}.Encode() resp, err := http.DefaultClient.Do(req) if err != nil { panic(err) } defer resp.Body.Close() var body struct { Results []struct { Title string `json:"title"` JournalName string `json:"journal_name"` PublishYear int `json:"publish_year"` } `json:"results"` } json.NewDecoder(resp.Body).Decode(&body) for _, paper := range body.Results { fmt.Printf("%s — %s %d\n", paper.Title, paper.JournalName, paper.PublishYear) } } ``` ```php PHP theme={null} response = HttpClient.newHttpClient() .send(request, HttpResponse.BodyHandlers.ofString()); // Parse response.body() with your preferred JSON library (Jackson, Gson, ...) System.out.println(response.body()); } } ``` You get back a JSON object with a `results` list of ranked papers. A `401` means the key is missing or wrong. ## What a response looks like Here's a real request and response on a paid plan, where open-access results carry full-text excerpts: ```bash theme={null} curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=what dose of creatine improves working memory" ``` ```json wrap theme={null} { "results": [ { "title": "Creatine and improvement in cognitive function: Evaluation of a health claim pursuant to article 13(5) of regulation (EC) No 1924/2006", "authors": ["Dominique Turck", "T. Bohn", "Montaña Cámara", "…"], "journal_name": "EFSA Journal", "publisher_name": "Wiley", "publish_year": 2024, "sjr_best_quartile": 1, "citation_count": 5, "doi": "10.2903/j.efsa.2024.9100", "url": "https://consensus.app/papers/creatine-and-improvement-in-cognitive-function-turck-bohn/7e703e828b815257b31687a25eceac9d/", "takeaway": "Creatine supplementation has not been established as a cause-and-effect relationship for improving cognitive function in one or more domains.", "full_text_chunks": [ "Section: Human intervention studies in healthy individuals conducted under normal conditions | Two studies had an initial loading phase of 20 g/day of creatine for 5 days, followed by 5 g/day … A significant improvement in working memory, as measured by the Backwards Digit Span test, was observed only in the study of shorter duration …", "Section: Weighing of the evidence | … the acute effect of creatine on working memory reported in some studies at doses of 20 g/day given for 5-7 days … was not observed at lower creatine doses, ranging from 2.2 to 14 g/day …" ] } ] } ``` Trimmed to one result with shortened excerpts. The `takeaway` gives you the paper's conclusion; the `full_text_chunks` give you the doses and study designs behind it, each labeled with the section of the paper it came from. ## How Consensus compares In blinded evaluations on real user queries, Consensus search returns more relevant papers than Google Scholar and than frontier models with web search. See [Benchmarks](/benchmarks). > [card] [More relevant than Google Scholar](/benchmarks#consensus-vs-google-scholar) > > 88.1% vs 81.8% precision on 500 real user queries, judged blind by PhD researchers. > [card] [Beats GPT-5.5 and Opus 4.8, 10x faster](/benchmarks#consensus-vs-frontier-models) > > More relevant top-20 papers at a median of 8.7 seconds, versus 75 to 103 seconds. ## What people build with the API > [card] [What you can build with the Consensus API](/use-cases) > > Teams use the API to power internal research tools, field-monitoring agents, and scientific agents. Phylo calls it directly from its agents to screen about 20,000 drug targets across hundreds of diseases. Browse the use cases. * **Internal tools:** paper search for scientific agents, field-monitoring agents, and literature tables that fill themselves. * **Research systems:** literature review pipelines, citation-grounded writing, and recurring evidence reports. ## Full text in the API Every search reaches inside the paper. Queries are matched against full text where available, and on paid plans each open-access result also carries the passages that answer your query, in `full_text_chunks`. To read a whole open-access paper, pass its `paper_id` to `GET /v1/papers/full_text` for its complete text as Markdown. > [card] [Full text in the API](/api-full-text) > > Excerpts and complete text: what comes back, which papers qualify, plans, and credits. ## Base URL and authentication All requests go to: ``` https://api.consensus.app ``` Every request must include your API key in the `x-api-key` header: ```http theme={null} x-api-key: YOUR_API_KEY ``` See [plans and access](/api-plans-and-access) for each plan's monthly calls, page sizes, features, and rate limits. ## Next steps > [card] [Plans and access](/api-plans-and-access) > > Monthly calls, page sizes, full-text access, rate limits, and errors by plan. > [card] [Search endpoint reference](/api-reference/search) > > Full parameter list, response schema, and interactive playground. > [card] [API on GitHub](https://github.com/Consensus-NLP/consensus-api) > > Client code, examples, and issues for the Consensus API. > [card] [MCP server](/consensus-mcp) > > Prefer a no-code path? Connect Consensus to Claude, ChatGPT, and other MCP clients. > [card] [Consensus website](https://consensus.app) > > Explore the Consensus research platform. # API plans and access Source: https://docs.consensus.app/api-plans-and-access What each Consensus plan includes on the API: monthly calls, papers per request, pagination, full-text excerpts, complete text, citations, and rate limits. The API is self-serve on every Consensus plan. Your plan sets how many calls you get each month, how many papers each request can return, and which features you can use. ## Get a key **1. Open the dashboard** Sign in to Consensus and go to [consensus.app/api-mcp](https://consensus.app/api-mcp/), or click your profile icon and choose **API & MCP Dashboard**. **2. Create a key** Click **Create API key**, name it, and copy it. If you already have a key, open **Keys and clients** and click **New key**. The full key is shown once. **3. Send it on every request** Pass the key in the `x-api-key` header. Requests without a valid key return `401`. ## What each plan includes | | Free | Pro | Teams | Deep | Enterprise | | - | - | - | - | - | - | | Included calls per month | 30 | 500 | 500 | 2,000 | Custom | | Max calls with additional usage | 30 | 1,000 | 1,000 | 10,000 | Custom | | Max papers per request (`page_size`) | 20 | 300 | 300 | 750 | 1,000 | | Pagination past the first page | No | Yes | Yes | Yes | Yes | | [Full-text excerpts](/api-full-text) | No | Yes | Yes | Yes | Yes | | [Complete text](/api-full-text#complete-text) of open-access papers\*\* | No | Yes | Yes | Yes | Yes | | Institutional full-text links (`institutional_access_url`)\* | No | Yes | Yes | Yes | Yes | | Citations and references (beta)\*\*\* | No | Yes | Yes | Yes | Yes | | Rate limit | 1 request/second | 1 request/second | 1 request/second | 1 request/second | Custom | \* Returned on the first 50 search results and on [Thread](/api-quickstart-threads) citations when you've selected your institution in Consensus settings. An organization key uses the institution of the person who created it. \*\* `GET /v1/papers/full_text`, at 1 credit per paper returned. \*\*\* `GET /v1beta/papers/citations` and `GET /v1beta/papers/references`, at 1 credit per 100 papers returned. Both are in beta, so their request and response schemas may change in the coming days. Every plan can use every search filter. Compare plans at [consensus.app/pricing](https://consensus.app/pricing). ## How calls are counted * **One call per 100 papers returned**, rounded up, with a minimum of one call per request. A request that returns 20 papers is 1 call; 150 papers is 2 calls. * **A [Thread](/api-quickstart-threads) is charged by the amount of work the agent does.** `mode` controls the effort: a `deep` run does more work than a `pro` run, so it costs more. A failed run costs nothing. While a Thread runs, its maximum possible cost counts as used. If that doesn't fit under your plan's limit, creating the Thread or adding a follow-up returns `429`. * **One credit per full text returned.** [Complete text](/api-full-text#credits) costs 1 credit for each paper whose Markdown comes back, and a credit counts the same as a call. Papers that come back blocked or not found cost nothing. A paper charged to you in the last 24 hours, through the API or MCP, normally comes back free. * **One credit per 100 citing or cited papers returned.** `GET /v1beta/papers/citations` and `GET /v1beta/papers/references` are charged by the papers on the page, rounded up, with a minimum of 1 credit per request. * **API and MCP share one monthly pool.** Searches from your API key and from your MCP connection draw down the same allowance. * **Usage resets on the 1st of each month.** Track it in the **Usage** tab of the API & MCP Dashboard. ### Additional usage Pro, Teams, and Deep can keep making calls past the included amount, up to the plan's maximum, at **\$0.05 per call**. Additional usage is off by default; turn it on in the API & MCP Dashboard. Free plans stop at 30 calls. ## Page size and pagination * **`page_size` defaults to 20.** A larger value is capped to your plan's maximum rather than rejected, so check the `page_size` echoed in the response. * **Pages are zero-indexed.** Requesting `page=1` or later needs a paid plan; on Free it returns `403`. * **A query returns at most 1,000 results in total.** `(page + 1) × page_size` can't exceed 1,000 on any plan. With `page_size=250`, pages 0–3 cover the full 1,000. To go wider, split the topic into narrower queries or tighter filters. ## Rate limits Every self-serve plan allows **one request per second** per account. A faster request returns `429` with a `retry-after` header; wait that long and retry. Enterprise accounts can have custom limits. The Threads endpoints follow the same one-request-per-second limit, except `GET /v1/threads/{thread_id}`, which allows up to three requests per second. That limit is shared across all of your account's Threads. While a Thread is running, poll it every `poll_after_seconds`: 5 seconds for `pro`, 20 for `deep`. `GET /v1/papers/full_text`, `GET /v1beta/papers/citations`, and `GET /v1beta/papers/references` each allow one request per second on every plan, Enterprise included. ## Errors | Status | Meaning | What to do | | - | - | - | | `401` | Missing, invalid, or revoked key | Check the `x-api-key` header, or create a new key | | `402` | Billing is past due | Resolve payment in your account; calls resume after | | `403` `feature_not_allowed` | Pagination or full-text excerpts on a Free plan | Upgrade to a paid plan | | `403` `feature_not_allowed` on a Thread with `skill` | Skills aren't enabled for your account yet | Leave `skill` unset | | `403` `feature_not_allowed` on `/v1/papers/full_text` | Complete text on a Free plan | Upgrade to a paid plan | | `403` `feature_not_allowed` on `/v1beta/papers/citations` or `/references` | Citations and references on a Free plan | Upgrade to a paid plan | | `409` on a follow-up | The Thread's latest interaction is still running | Poll the Thread until its status is `idle` or `failed`, then retry | | `429` "used all included searches" | Monthly calls used up | Turn on additional usage, upgrade, or wait for the 1st | | `429` "Too many requests" | More than one request per second | Wait for `retry-after`, then retry | ## Enterprise Enterprise adds custom call volumes and rate limits and up to 1,000 papers per request. An Enterprise contract covers search only: [complete text](/api-full-text#complete-text), citations and references, and [Threads](/api-quickstart-threads) are billed per credit at the standard rate, on top of the contract. Enterprise keys authenticate the same way as every other key: send them in the `x-api-key` header on requests to `api.consensus.app`. > [!NOTE] > > `Authorization: Bearer` is for the MCP server (`mcp.consensus.app/mcp`), not the REST API. See [MCP API key access](/mcp-plans-and-access#api-key-access). For Enterprise access, [contact sales](https://consensus.app/home/contact/sales/). Setting Consensus up behind an AI gateway or in Microsoft 365 Copilot? See the [Enterprise guides](/enterprise/overview). ## Related > [card] [Get started](/api-get-started) > > Your first request, and a real response with full-text excerpts. > [card] [MCP plans and access](/mcp-plans-and-access) > > The same plans on the MCP server, with its per-plan result and field limits. # Quick start: Search Source: https://docs.consensus.app/api-quickstart-search Find ranked papers for a query in one request, and see when search is the right Consensus API endpoint. `GET /v1/search` returns ranked papers for a query in a single request. Use it when your code or your own model needs the papers themselves: titles, journals, citation counts, links, and full-text excerpts. ## When to use search Search is the right fit when you want the papers and plan to do the reasoning yourself. * **Lookups:** find the papers on a specific question, author, or method. * **Your own pipeline:** feed ranked papers and full-text excerpts into your own LLM, agent, or scoring step. * **High volume:** screen many queries or targets quickly. Each request costs one call per 100 papers returned. * **Structured filtering:** narrow by study design, date, journal quality, sample size, field, or country. Want Consensus to read the papers and write a cited answer for you instead? Use [Threads](/api-quickstart-threads). ## Make a request Store your key in an environment variable named `CONSENSUS_API_KEY`, then send a `query`: ```bash theme={null} curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=does creatine improve working memory" \ --data-urlencode "study_types=rct,meta-analysis" \ --data-urlencode "year_min=2015" ``` Each result in `results` carries the paper's `title`, `authors`, `journal_name`, `publish_year`, `citation_count`, `doi`, a Consensus `url`, and a one-sentence `takeaway`, plus its `paper_id` and `full_text_available`. On a paid plan, open-access results also carry `full_text_chunks`: the passages from the paper's full text that answer your query. ## Go further * **Full-text excerpts:** returned on paid plans for open-access papers. See [Full text in the API](/api-full-text). * **Complete text:** pass the `paper_id` of results where `full_text_available` is `true` to `GET /v1/papers/full_text` to read the whole paper as Markdown. Paid plans. See [complete text](/api-full-text#complete-text). * **Institutional full-text links:** on a paid plan, if you've selected your institution in Consensus settings, the first 50 results can include an `institutional_access_url` that opens the full text through your library. * **More results:** raise `page_size` and page through with `page`. Your plan sets the maximum. See [plans and access](/api-plans-and-access). * **Every parameter:** see the [search endpoint reference](/api-reference/search). ## Use cases > [card] [Abstract screening](/use-cases/abstract-screening) > > Screen candidate papers against inclusion criteria at scale. > [card] [Citation verification](/use-cases/citation-verification) > > Check that a claim is backed by the paper it cites. > [card] [Extract values from full text](/use-cases/extract-values-from-full-text) > > Pull doses, effect sizes, and sample sizes from full-text excerpts. > [card] [Scheduled research pipeline](/use-cases/scheduled-research-pipeline) > > Monitor a field on a schedule and report what's new. # Quick start: Threads Source: https://docs.consensus.app/api-quickstart-threads Delegate a research question to the Consensus research agent and get back a cited answer, using the Threads endpoints of the Consensus API. The Threads endpoints hand a research question to the Consensus research agent. The agent searches the literature, reads the papers, and writes a markdown answer with citations. You create a Thread, poll it until the answer is ready, and add follow-ups that keep the earlier context. ## Search vs Threads Search returns papers and leaves the reasoning to your code. A Thread hands the whole research task to an agent, which runs many steps and writes the answer for you. | | Search | Threads | | - | - | - | | **You get back** | A list of papers with abstracts and metadata | A list of papers, a written analysis, a reasoning trace (with `include_trace=true`), and citations with grounding quotes | | **Functionality** | Semantic paper search with filters | Search, citation crawls, author search, finding specific papers, reading full papers, searching attached collections, and reading and writing your Library collections | | **Steps per request** | One | Many | | **Timing** | Synchronous | Asynchronous | | **Follow-ups** | Each request stands alone | Ask follow-up questions that build on earlier answers | Need the raw papers for your own pipeline, or a quick lookup? Use [Search](/api-quickstart-search). ## What you can ask a Thread Each of these needs several research steps in one request, which Search can't do. Click a card for the request body. > [card] [Crawl citations](#crawl-citations-from-a-seed-paper) > > Trace what a paper built on and the work that followed it. > [card] [Read specific papers](#read-and-compare-specific-papers) > > Compare the methods and findings of papers you choose. > [card] [Research an author](#research-an-author) > > Summarize an author's work on a topic and how it changed. > [card] [Run deep research](#run-a-deep-literature-review) > > Get a structured literature review built from about 50 papers. > [card] [Build on an earlier answer](#build-on-an-earlier-answer) > > Narrow or extend the last answer without starting over. ## How it works Threads are asynchronous. A `pro` run usually finishes in under a minute, and a `deep` run takes a few minutes. **1. Create a Thread** `POST /v1/threads` starts the run and returns a `thread_id` right away, with `status: "running"`. **2. Poll until it's done** `GET /v1/threads/{thread_id}` until `status` is `idle` (answer ready) or `failed`. Wait `poll_after_seconds` between polls: 5 seconds for `pro`, 20 for `deep`. **3. Ask a follow-up** `POST /v1/threads/{thread_id}/interactions` continues the Thread. Poll again for the new answer. ## Create a Thread Send the complete research brief in `input_message`. The agent doesn't ask clarifying questions, so include the whole task in one message. ```bash theme={null} curl -X POST "https://api.consensus.app/v1/threads" \ -H "x-api-key: $CONSENSUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "input_message": "What dose of creatine improves working memory in healthy adults, and how strong is the evidence?", "mode": "pro", "filters": { "year_min": 2010, "human": true } }' ``` ```json theme={null} { "thread_id": "", "interaction_id": "", "title": "Creatine dose and working memory in healthy adults", "status": "running", "mode": "pro", "url": "https://consensus.app/search/creatine-dose-and-working-memory-in-healthy-adults//", "poll_after_seconds": 5, "max_wait_seconds": 180 } ``` * **`mode`:** controls how much effort the agent spends. `pro` (the default) fits most questions. `deep` reads more papers and writes a longer, structured report. * **`filters`:** the same filters as `GET /v1/search`, such as `year_min`, `study_types`, `human`, and `sjr_max`. They apply to every search the agent runs. * **`attachments`:** focus the research on up to 50 `paper_ids` or 10 Library `collection_ids`. A collection's id is the number in its web URL, `consensus.app/library/{id}/`. * **`skill`:** run a named research workflow, a [Skill](/core-features/skills), instead of the agent's open-ended plan: `literature-review`, `citation-snowballing`, `find-citations`, or `verify-citations`. It requires `mode` set to `deep`. Leave it unset for normal research. ## Poll for the answer ```bash theme={null} curl "https://api.consensus.app/v1/threads/" -H "x-api-key: $CONSENSUS_API_KEY" ``` While the run is going, `status` is `running`. When it's `idle`, the latest interaction carries the answer: ```json theme={null} { "thread_id": "", "title": "Creatine dose and working memory in healthy adults", "status": "idle", "url": "https://consensus.app/search/creatine-dose-and-working-memory-in-healthy-adults//", "interactions": [ { "interaction_id": "", "status": "idle", "mode": "pro", "response": "Short-term loading at about 20 g/day improved working memory in some trials [1.1] …", "citations": [ { "n": 1, "paper_id": "", "title": "Creatine and improvement in cognitive function: Evaluation of a health claim …", "year": 2024, "journal": "EFSA Journal", "url": "https://consensus.app/papers/…", "snippets": [ { "marker": "1.1", "quote": "…", "section": "Weighing of the evidence" } ] } ], "stats": { "retrieved": 120, "screened": 40, "included": 18 } } ] } ``` Illustrative and trimmed. In `response`, a marker like `[1.1]` cites quote 1 from paper 1, and a bare `[1]` cites the paper without a quote. Each `citations[].snippets[].marker` matches a marker in the text. Quotes can come from the full text of any paper the analysis cites, paywalled papers from Consensus's [publisher partners](/get-started/full-text-access) included. * **How long to wait:** stop polling after `max_wait_seconds`, about 3 minutes for `pro` and 10 minutes for `deep`. * **Agent steps:** add `include_trace=true` to get `agent_trace`, a plain-English list of the steps the agent took. It's written for display, so don't parse it. * **Skill:** an interaction that ran a [Skill](/core-features/skills) carries `skill`, such as `literature-review` or `verify-citations`: the one requested, otherwise the one the agent chose itself. It's omitted when none applies. * **Earlier turns:** only the latest interaction carries `response` and `citations`. Pass `interaction_id` to read an earlier one. * **Institutional full-text links:** on a paid plan, if you've selected your institution in Consensus settings, citations can include an `institutional_access_url` that opens the full text through your library. Only the first 50 distinct DOIs are looked up. ## Ask a follow-up ```bash theme={null} curl -X POST "https://api.consensus.app/v1/threads//interactions" \ -H "x-api-key: $CONSENSUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "input_message": "Do the results differ for adults over 60?" }' ``` The agent already has the earlier turns, so don't restate them. A follow-up sent while the latest interaction is still `running` returns `409`. Wait until the Thread is `idle` or `failed`, then send it. To find earlier Threads, `GET /v1/threads` lists yours newest first. Pass `query` to match on title. ## Work in your Library With a personal API key, the agent can read and write your [Library](/core-features/library) as part of a Thread. Ask in plain language in `input_message`, and name collections the way they appear in Consensus. If two collections share a name, also name the parent collection, for example "the Methods collection inside Sleep". There are no separate Library endpoints. | Action | Example `input_message` | | - | - | | List your collections, with their paper counts and your role on each | "List all of my Library collections and how many papers each has." | | List the papers in a collection, filtered by publication year, date added, or a title, author, or journal match | "Which papers in my Sleep collection were published before 2020?" | | Search within your collections | "What do the papers in my Sleep collection say about magnesium?" | | Create a collection, at the top level or inside another one | "Create a collection called Magnesium inside my Sleep collection." | | Save papers the agent found into a collection | "Find 10 RCTs on magnesium and sleep quality and save them to my Sleep collection." | | Rename a collection | "Rename my Sleep collection to Sleep and Supplements." | * **Saves don't pause for confirmation.** The agent saves papers as soon as you ask, and the answer says how many it added and how many were already there. * **No surprise collections.** The agent creates a collection only when you ask for one. If you name a collection that doesn't exist, it reports it as not found instead of creating it. * **Removing, deleting, and moving are web-only.** These need a confirmation step, so they're available only in the Consensus web app. Over the API, the agent tells you to make the change there. * **Listings stop at 100 collections.** In a larger Library, the rest are summarized as "and N more", and you can't page through them. * **Personal keys only.** An organization API key has no Library, so its Threads can't list or change collections. The Thread's `response` reports what changed. ## Example requests Each example shows the request body. Send it the same way as in [Create a Thread](#create-a-thread), then poll for the answer. ### Crawl citations from a seed paper The agent follows the citation graph backward and forward from a paper you name. ```json theme={null} { "input_message": "Start from Cong et al. 2013, the first report of CRISPR-Cas9 genome editing in mammalian cells. Trace the earlier work it built on and the most-cited papers that followed it, and summarize how the method evolved." } ``` ### Read and compare specific papers Attach papers with `attachments.paper_ids`, and the agent reads them before it answers. Paper IDs come from `citations[].paper_id` in an earlier Thread. ```json theme={null} { "input_message": "Compare the methods, sample sizes, and effect sizes across these papers, and explain where their conclusions disagree.", "attachments": { "paper_ids": ["", "", ""] } } ``` ### Research an author The agent looks up papers by a named author, then synthesizes them. ```json theme={null} { "input_message": "Summarize Walter Willett's research on dietary fat and heart disease, and how his conclusions changed over time." } ``` ### Run a deep literature review Set `mode` to `deep` for a structured report built from about 50 papers. A `deep` run takes up to about 10 minutes and requires a paid plan. ```json theme={null} { "input_message": "Write a literature review on GLP-1 receptor agonists and cardiovascular outcomes. Cover mechanisms, the major trials, safety signals, and open questions.", "mode": "deep", "filters": { "year_min": 2015 } } ``` ### Run a skill Set `skill` to run a named workflow, on a new Thread or a follow-up. Skills run in `deep` mode, so set `mode` to `deep` too. `find-citations` and `verify-citations` work on a draft, so put the draft text in `input_message`. ```json theme={null} { "input_message": "Check whether the citations in this paragraph support its claims: …", "mode": "deep", "skill": "verify-citations" } ``` A request with `skill` is refused with `422` if `mode` isn't `deep`, if you use an organization API key, or if the skill needs input the request didn't include. If skills aren't enabled for your account yet, it returns `403` `feature_not_allowed`. ### Save papers to a collection The agent runs the research, then saves the papers it found into a collection you name. Ask for a new collection explicitly if you want one. ```json theme={null} { "input_message": "Find 10 randomized controlled trials on magnesium supplementation and sleep quality in adults, summarize their findings, and save them to a new collection called Magnesium and Sleep." } ``` ### Audit and tidy a collection The agent lists a collection's papers with filters, reads them, and renames the collection. ```json theme={null} { "input_message": "List the papers in my Sleep collection published before 2015, summarize which of their conclusions newer work has overturned, then rename the collection to Sleep - foundational." } ``` ### Build on an earlier answer Send a follow-up to `POST /v1/threads/{thread_id}/interactions`. The agent reads the earlier turns, so you can refer to them directly. ```json theme={null} { "input_message": "Of the trials you cited, keep only the randomized controlled trials with more than 100 participants. Does the conclusion change?" } ``` ## Good to know * **Cost:** a Thread draws from the same monthly calls as search, and it's charged by the amount of work the agent does. `mode` controls the effort, so a `deep` run costs more than a `pro` run. A failed run costs nothing. See [plans and access](/api-plans-and-access#how-calls-are-counted). * **Rate limits:** one request per second, except polling a Thread, which allows three per second. * **Don't retry blindly on `5xx`.** The Thread or follow-up may already have been created. List your Threads or poll the Thread before you send it again. ## Endpoint reference > [card] [Start a Thread](/api-reference/start-a-thread) > > `POST /v1/threads` > [card] [Poll a Thread](/api-reference/poll-a-thread) > > `GET /v1/threads/{thread_id}` > [card] [Add a follow-up](/api-reference/add-a-follow-up) > > `POST /v1/threads/{thread_id}/interactions` > [card] [List historical Threads](/api-reference/list-historical-threads) > > `GET /v1/threads` # Add a follow-up Source: https://docs.consensus.app/api-reference/add-a-follow-up /openapi.json post /v1/threads/{thread_id}/interactions Continue the research conversation with full context of the earlier turns. Rate limit: 1 request/second. Refused with 409 while the Thread's latest interaction is still running, so wait until status is `idle` or `failed` first. Returns a new `interaction_id` immediately; poll `GET /v1/threads/{thread_id}` until status is `idle`. A 5xx means the request may or may not have taken effect (no idempotency key yet). Set `skill` (with `mode` `deep`) to run a named research workflow for this follow-up. # Get citations Source: https://docs.consensus.app/api-reference/citations-beta/get-citations /openapi.json get /v1beta/papers/citations Find the papers that cite a set of seed papers, up to 20 seeds in one request. Each paper comes back in the same shape as a search result. **Which seeds:** pass `paper_ids` (the `paper_id` on search, Thread, citations or references results), `dois`, or both. 1 to 20 in total. DOIs that match no Consensus paper come back in `unmatched_dois`. **Several seeds:** each paper carries `seeds_touched`, how many of the seeds it cites. With the default `sort`, papers that cite more of the seeds rank first. **Filters and order:** `year_min`, `year_max` and `cite_min` narrow the results. `sort` is `recency_weighted` (the default), `most_cited` or `most_recent`. **Paging:** `page` starts at 0 and `page_size` defaults to 20, up to your plan's maximum. Request `next_page` until `is_end` is true. Paging reaches the first 10,000 papers. **Cost:** 1 credit per 100 papers returned, rounded up, with a minimum of 1 per request. `credits_charged` shows what the request used. **Access:** paid plans only. A `403` means citations aren't available to your key. **Rate limit:** 1 request per second. > [!WARNING] > > This endpoint is in beta, so its request and response schemas may change in the coming days. # Get references Source: https://docs.consensus.app/api-reference/citations-beta/get-references /openapi.json get /v1beta/papers/references Find the papers that a set of seed papers cites, up to 20 seeds in one request. Each paper comes back in the same shape as a search result. **Which seeds:** pass `paper_ids` (the `paper_id` on search, Thread, citations or references results), `dois`, or both. 1 to 20 in total. DOIs that match no Consensus paper come back in `unmatched_dois`. **Several seeds:** each paper carries `seeds_touched`, how many of the seeds cite it. With the default `sort`, papers cited by more of the seeds rank first. **Filters and order:** `year_min`, `year_max` and `cite_min` narrow the results. `sort` is `recency_weighted` (the default), `most_cited` or `most_recent`. **Paging:** `page` starts at 0 and `page_size` defaults to 20, up to your plan's maximum. Request `next_page` until `is_end` is true. Paging reaches the first 10,000 papers. **Cost:** 1 credit per 100 papers returned, rounded up, with a minimum of 1 per request. `credits_charged` shows what the request used. **Access:** paid plans only. A `403` means citations aren't available to your key. **Rate limit:** 1 request per second. > [!WARNING] > > This endpoint is in beta, so its request and response schemas may change in the coming days. # Get full text Source: https://docs.consensus.app/api-reference/get-full-text /openapi.json get /v1/papers/full_text Read the full text of up to 20 open-access papers in one request. Each paper comes back as Markdown, split into sections. **Which papers:** pass `paper_ids` (the `paper_id` on search or Thread results), `dois`, or both. 1 to 20 in total. **Each paper's `status`:** - `available`: the full text is in the response. - `blocked`: Consensus can't return this paper's text. It isn't open access, it was retracted, or the copy Consensus holds isn't the open-access version. - `no_full_text`: Consensus has no text for this paper. - `not_found`: no paper matches that id or DOI. **Cost:** 1 credit per `available` paper. Every other status is not charged. Reading a paper again within 24 hours of being charged for it is also free; `free_reread_until` tells you until when. **Access:** paid plans only. A `403` means full text isn't available to your key. **Rate limit:** 1 request per second. # List historical Threads Source: https://docs.consensus.app/api-reference/list-historical-threads /openapi.json get /v1/threads The caller's prior Threads, newest first, optionally filtered by a case-insensitive title substring. Rate limit: 1 request/second. On this list, status may be absent for Threads inactive past the status window and title is substituted when the Thread has none; GET the Thread for authoritative status and title. # Poll a Thread Source: https://docs.consensus.app/api-reference/poll-a-thread /openapi.json get /v1/threads/{thread_id} The polling half of create-then-poll. While status is `running`, wait `poll_after_seconds` from the dispatch response between polls (5 for pro, 20 for deep); faster polling is rate-limited (hard cap 3/s per account, shared across all of its Threads). Readable for any Thread the key can access, including ones shared with the caller, not only Threads it created. Only the target interaction (`interaction_id`, or the latest) carries response/citations; earlier turns are metadata unless requested explicitly. # Quick search Source: https://docs.consensus.app/api-reference/quick-search /openapi.json get /v1/quick_search This endpoint is deprecated. Use `GET /v1/search` instead. It will be removed on 2027-02-07. # Search Source: https://docs.consensus.app/api-reference/search /openapi.json get /v1/search The supported endpoint for searching academic papers. Use this instead of the legacy `/v1/quick_search` endpoint, which is deprecated and will be removed on 2027-02-07. # Start a Thread Source: https://docs.consensus.app/api-reference/start-a-thread /openapi.json post /v1/threads Delegate a research task to the Consensus research agent and get a `thread_id` back immediately. Rate limit: 1 request/second. The run is asynchronous: poll `GET /v1/threads/{thread_id}` until status is `idle`. A 5xx means the request may or may not have taken effect: there is no idempotency key yet, so a blind retry can create a second Thread. Set `skill` (with `mode` `deep`) to run a named research workflow such as a literature review. # What you can build with the Consensus API Source: https://docs.consensus.app/api-use-cases # How Consensus compares Source: https://docs.consensus.app/benchmarks Blinded evaluations of Consensus search on real user queries: more relevant papers than Google Scholar (88.1% vs 81.8% precision), and more relevant than GPT-5.5 and Opus 4.8 with web search at a median of 8.7 seconds versus 75 to 103 seconds. Consensus publishes head-to-head evaluations of its search on real, anonymized user queries. Independent PhD-level researchers score every result blind to which system returned it. This page summarizes each evaluation and links to the full write-up. *Last updated: October 27, 2026.* ## At a glance | Evaluation | Published | Queries | Judges | Result | | - | - | - | - | - | | [Consensus vs frontier models](#consensus-vs-frontier-models) | August 4, 2026 | 100 | 4 PhD researchers, plus LLM-as-a-judge | Consensus Pro was 4.5% more precise than GPT-5.5 and 13.2% more precise than Opus 4.8 in the top 20, with an 8.7 s median latency versus 74.8 s and 102.8 s | | [Consensus vs Google Scholar](#consensus-vs-google-scholar) | December 5, 2025 | 500 | 3 PhD researchers | Consensus Quick search reached 88.1% precision in the top 10 versus 81.8% for Google Scholar (7.4% better), and 10.3% better DCG | Precision is the share of returned papers that judges scored 3 or 4 on a 0 to 4 relevance scale, meaning a researcher would consider them genuinely relevant to the query. ## Consensus vs frontier models Consensus Pro returned more relevant papers than GPT-5.5 and Opus 4.8, each with web search, on 100 real user queries. In the top 20 results it was 4.5% more precise than GPT-5.5 and 13.2% more precise than Opus 4.8. It was also about 10x faster: a median of 8.7 seconds end to end, against 74.8 seconds for GPT-5.5 and 102.8 seconds for Opus 4.8. | Metric | Consensus Pro vs GPT-5.5 | Consensus Pro vs Opus 4.8 | | - | - | - | | Precision, top 10 | 1.8% better | 11.6% better | | Precision, top 20 | 4.5% better | 13.2% better | | Median latency | 8.6x faster (8.7 s vs 74.8 s) | 11.8x faster (8.7 s vs 102.8 s) | | p75 latency | 3.8x faster | 5.1x faster | Query by query, Consensus had the most relevant results of the three systems on 59 of the 100 queries. GPT-5.5 led on 22 and Opus 4.8 on 6, with 11 ties. All three systems lose precision deeper in the list, but the general models lose it faster: GPT-5.5 drops 10.3 points between rank 10 and rank 20, and Consensus drops 8.7. **How it was run** * **Systems**: Consensus Pro, the agentic search mode that decomposes a query, runs multiple retrievals against the Consensus index, and re-ranks. GPT-5.5 and Opus 4.8 each ran with web search enabled and received the same query. * **Queries**: 100 randomly sampled queries from real Consensus users, anonymized and scrubbed of PII. They range from keyword lookups and metadata queries to natural-language research questions. * **Judging**: four independent PhD-level researchers scored every query–paper pair from 0 (irrelevant) to 4 (perfectly relevant), blind to the source system. LLM-as-a-judge across several frontier models gave a second, independent signal. * **Latency**: wall-clock time from query submission to a complete set of results. **Limitations**: judges saw titles and abstracts only, because some systems don't expose full text. Scores reflect textual relevance, not recency, citation count, or journal quality. The evaluation measured time, not token or dollar cost. > [card] [Read the full evaluation](https://consensus.app/home/blog/consensus-outperforms-frontier-models-in-scientific-search/) > > Consensus Beats Claude and ChatGPT in Academic Search, on the Consensus blog. ## Consensus vs Google Scholar Consensus Quick search returned more relevant papers than Google Scholar on 500 real user queries. In the top 10 results, 88.1% of Consensus papers were relevant versus 81.8% for Google Scholar, 7.4% better. Consensus also ranked relevant papers higher, with 10.3% better DCG, and had higher average relevance at every rank in the list. | Metric | Consensus | Google Scholar | Difference | | - | - | - | - | | Precision, top 10 | 88.1% | 81.8% | 7.4% better | | DCG, top 10 | 45.16 | 40.73 | 10.3% better | **How it was run** * **Systems**: Consensus Quick search, a single-pass search, against Google Scholar. Consensus Pro and Deep modes, which run iterative multi-query searches, weren't tested. * **Queries**: 500 randomly sampled queries from real Consensus users, anonymized and scrubbed of PII, from keywords and metadata queries to natural-language questions. * **Judging**: the top 10 results from each system made about 10,000 query–paper pairs. Three independent PhD researchers, each with at least a decade of experience assessing research, scored every pair from 0 to 4, blind to the source system. **Limitations**: judges saw titles and abstracts only. Scores reflect textual relevance, not recency, citation count, or journal quality. > [card] [Read the full evaluation](https://consensus.app/home/blog/consensus-outperforms-google-scholar-for-academic-search-retrieval/) > > Consensus Outperforms Google Scholar for Academic Search Retrieval, on the Consensus blog. ## Use Consensus search The same search is available in the web app, over the [REST API](/api-get-started), and as an [MCP server](/consensus-mcp) for ChatGPT, Claude, Perplexity, Cursor, and other AI assistants. To see how ranking works, read [Paper search](/get-started/how-search-works). # Get started with Consensus MCP Source: https://docs.consensus.app/consensus-mcp Search 400M+ scholarly articles, including 350M peer-reviewed papers and all of PubMed, from Claude, ChatGPT, Cursor, Microsoft 365 Copilot and other MCP clients. One hosted server covers every field of science, so you don't need separate PubMed, OpenAlex or Semantic Scholar servers. Your assistant gets ranked papers it can filter by study design, sample size and journal quality, and cites real studies in its answers. Connect your client below. You'll be searching in under 2 minutes. > [card] [What you can build with the Consensus MCP](/use-cases) > > MCP is the fastest way to give an assistant or agent paper search. Owkin plugged Consensus MCP into its agentic AI scientist and went from curated PubMed abstracts to 25x the paper coverage in days. Browse the use cases. **Server URL** ```text theme={null} https://mcp.consensus.app/mcp ``` > [card] [Claude](https://claude.ai/directory/65247229-f0c7-49df-9044-fcbb8b3894c6) > > Add Consensus from the Claude connectors directory. > [card] [ChatGPT](https://chatgpt.com/plugins/plugin_asdk_app_6943e6f4a928819195962de16fb9ffe4) > > Add Consensus from the ChatGPT app directory. > [card] [Microsoft 365 Copilot](https://marketplace.microsoft.com/en-us/product/WA200011861) > > Add Consensus for Copilot from the Microsoft Marketplace. > [card] [Perplexity](https://www.perplexity.ai/computer/connectors?query=consensus\&connector=consensus) > > Add Consensus from the Perplexity Computer connectors store. ## How Consensus compares In blinded evaluations on real user queries, Consensus search returns more relevant papers than Google Scholar and than frontier models with web search. See [Benchmarks](/benchmarks). > [card] [More relevant than Google Scholar](/benchmarks#consensus-vs-google-scholar) > > 88.1% vs 81.8% precision on 500 real user queries, judged blind by PhD researchers. > [card] [Beats GPT-5.5 and Opus 4.8, 10x faster](/benchmarks#consensus-vs-frontier-models) > > More relevant top-20 papers at a median of 8.7 seconds, versus 75 to 103 seconds. ## Connect Your Client > [accordion] Claude Desktop > > **One-click:** open [Consensus in the Claude directory](https://claude.ai/directory/65247229-f0c7-49df-9044-fcbb8b3894c6) and click **Connect**. Or connect manually: > > **1. Open Settings** > > Open [claude.ai](https://claude.ai) or the Claude desktop app and go to **Settings**. > > **2. Browse connectors** > > Select **Connectors**, then **Browse connectors**. > > **3. Find Consensus** > > Search for **Consensus** and click **Connect**. > > **4. Sign in** > > Approve the sign-in with your Consensus account, or continue without one at reduced limits. > > **5. Start searching** > > Ask a research question in any chat. Consensus appears in the tools menu. > [accordion] Claude Code > > ```bash theme={null} > claude mcp add --transport http consensus https://mcp.consensus.app/mcp > ``` > > OAuth will open in your browser automatically on first use. You can also trigger it manually with the `/mcp` command inside Claude Code. > [accordion] ChatGPT > > **One-click:** open [Consensus in the ChatGPT app directory](https://chatgpt.com/plugins/plugin_asdk_app_6943e6f4a928819195962de16fb9ffe4) and click **Connect**. Or connect manually: > > **1. Open ChatGPT** > > Open [chatgpt.com](https://chatgpt.com) and sign in. > > **2. Find Consensus** > > Open **Apps** from the sidebar, search for **Consensus**, and click **Connect**. > > **3. Sign in** > > Sign in with your Consensus account when prompted. > > **4. Start searching** > > Ask a research question, or call it by name: "Use Consensus to find RCTs on…" > > Search results display in an interactive widget showing top papers with citations and direct links. > > > [!NOTE] > > > > Consensus also works with ChatGPT Deep Research. See [Using Consensus with ChatGPT Deep Research](#using-consensus-with-chatgpt-deep-research) below. > [accordion] Codex > > ```bash theme={null} > codex mcp add consensus --url https://mcp.consensus.app/mcp > codex mcp login consensus > ``` > > Or add to `~/.codex/config.toml`: > > ```toml theme={null} > [mcp_servers.consensus] > url = "https://mcp.consensus.app/mcp" > ``` > [accordion] Cursor > > **One-click:** [Install in Cursor](https://cursor.com/en/install-mcp?name=consensus\&config=eyJ1cmwiOiJodHRwczovL21jcC5jb25zZW5zdXMuYXBwL21jcCJ9). Or add to `~/.cursor/mcp.json`: > > ```json theme={null} > { > "mcpServers": { > "consensus": { > "command": "npx", > "args": ["-y", "mcp-remote@latest", "https://mcp.consensus.app/mcp"] > } > } > } > ``` > [accordion] VS Code (GitHub Copilot) > > Add to your MCP configuration: > > ```json theme={null} > { > "servers": { > "consensus": { > "url": "https://mcp.consensus.app/mcp" > } > } > } > ``` > [accordion] Windsurf > > Add to your MCP configuration: > > ```json theme={null} > { > "mcpServers": { > "consensus": { > "serverUrl": "https://mcp.consensus.app/mcp" > } > } > } > ``` > [accordion] Gemini Spark > > Add Consensus as a custom app for Gemini Spark: > > **1. Open Connected Apps** > > Go to [gemini.google.com](https://gemini.google.com) and sign in with a personal Google Account, then click **Settings & help** → **Connected Apps**. > > **2. Add a custom app** > > Under **Custom apps for Spark**, click **Add a custom app** and paste the server URL: > > ```text theme={null} > https://mcp.consensus.app/mcp > ``` > > **3. Approve the connection** > > Click **Next** and follow the prompts to finish OAuth. If your client doesn't support Dynamic Client Registration, expand **Advanced features** and enter your credentials. > > **4. Start searching** > > Ask Gemini Spark a research question and it can call Consensus tools. > > > [!NOTE] > > > > Requires access to Gemini Spark, a personal Google Account, Keep Activity turned on, and is currently available in the US. Workspace and school accounts are not supported yet. > [accordion] Perplexity > > **1. Open the Consensus connector** > > Go to [Consensus in the Perplexity connectors store](https://www.perplexity.ai/computer/connectors?query=consensus\&connector=consensus). > > **2. Add the connector** > > Click **+ Add Connector**. > > **3. Sign in with Consensus** > > Complete the OAuth sign-in with your Consensus account. You need a [Consensus account](https://consensus.app/sign-up), which is free to create. > > **4. Confirm the connection** > > Click **Connect** on the OAuth confirmation screen. > > **5. Open Computer** > > Click **Computer**. > > **6. Ask a research question** > > Say "Use Consensus to..." followed by a scientific question, for example: "Use Consensus to identify the pros and cons of GLP-1s." > [accordion] Microsoft 365 Copilot > > In Microsoft 365 Copilot, Consensus is a connector your admin sets up for the tenant. Once it's there, connect it yourself under **Sources**. > > [YouTube video player](https://www.youtube.com/embed/g2_WBlJupx4) > > **1. Open Copilot Chat settings** > > In Microsoft 365 Copilot, select the **gear icon** at the bottom of the left sidebar, then **Settings**, then **Sources**. > > **2. Connect Consensus** > > Find **Consensus for Copilot** and select **Connect**, then sign in with your Consensus account. > > **3. Ask a research question** > > Type a research question like 'What does the research say about creatine and cognitive function?' and Copilot will call Consensus to find relevant papers. > > > [!NOTE] > > > > Don't see Consensus under Sources? Your admin needs to add it first. Send them the [Microsoft 365 Copilot admin guide](/enterprise/microsoft-365-copilot). A [Consensus account](https://consensus.app/sign-up) is required and free to create. > [accordion] Microsoft 365 Copilot Studio > > This allows you to add Consensus as a custom tool in Copilot studio so it can be used with agents or other automations you have have. This type of authentication will  > > **1. Open custom tools** > > Go to Copilot studio and find the 'tools' tab on the left hand side. > > ![Clean Shot 2026 08 31 At 17 17 16@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.17.16@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=b5f6a6010734943bb41b7d8351b2e952) > > **2. Create 'Create new tool'** > > ![Clean Shot 2026 08 31 At 17 19 50@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.19.50@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=cf3823beb65f9b462ad5c1f658e9b6ba) > > Click 'new tool' to add a custom tool > > **3. Click 'Model context protocol'** > > ![Clean Shot 2026 08 31 At 17 21 06@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.21.06@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=7bb99d2343cd51fa74f3ace9e886aabe) > > You can also use the API to do this. But to use the MCP, click 'model context protocol'. > > **4. Fill in the connector information** > > ![Clean Shot 2026 08 31 At 17 27 17@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.27.17@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=557f2599cb9066c2bb71177b2bc4b402) > > **Server name** > > Consensus > > **Server Description** > > Consensus is the go-to MCP for academic research. Search 400M+ scholarly articles, including 350M peer-reviewed papers, from PubMed, Semantic Scholar, and ArXiv — filter by study type (RCTs, meta-analyses, systematic reviews), journal quality, sample size, and more. Get structured results with citations, study takeaways, and direct links, or run pre-built workflows for literature reviews, grant research, and evidence synthesis. > > **Server URL** > > [https://mcp.consensus.app/mcp](https://mcp.consensus.app/mcp) > > **Authentication** > > OAuth 2.0 > > **Type** > > Dynamic Discovery > > **5. Save and use!** > > Thats it! This custom tool can now be used and shared with other team members. Each team member will be able to connect to their own Consensus account when they use the tool. > > > [!NOTE] > > > > A [Consensus account](https://consensus.app/sign-up) is required. It is free to create. > [accordion] API key (Bearer token) > > Use an API key for clients without OAuth, headless agents, and server-side automations. Send it in the `Authorization` header on every request to `https://mcp.consensus.app/mcp`: > > ```http theme={null} > Authorization: Bearer YOUR_API_KEY > ``` > > Create a key from your profile icon under **API & MCP Dashboard** → **Keys and clients** → **New key**. The key gets its owner's plan limits and draws from the same monthly pool as the API. See [MCP plans and access](/mcp-plans-and-access). > > **JSON config** (Cursor, VS Code, and most clients that accept headers): > > ```json theme={null} > { > "mcpServers": { > "consensus": { > "url": "https://mcp.consensus.app/mcp", > "headers": { > "Authorization": "Bearer YOUR_API_KEY" > } > } > } > } > ``` > > **Claude Code:** > > ```bash theme={null} > claude mcp add --transport http consensus https://mcp.consensus.app/mcp --header "Authorization: Bearer $CONSENSUS_API_KEY" > ``` > > > [!WARNING] > > > > An invalid or revoked key fails the search with a tool error: "Your API key is invalid or expired." Create a new key from the **API & MCP Dashboard** and update your client config. > > Keep the key server-side. Don't paste it into prompts or commit it to a repo. > [accordion] Other MCP clients > > Any client that supports remote MCP servers (Streamable HTTP) works. Point it at `https://mcp.consensus.app/mcp`. If your client supports OAuth, authentication is handled automatically. Pick your client for step-by-step setup. [The Consensus MCP & Claude Connector](https://www.youtube.com/embed/gcMel2guYE8?start=4) Watch: connect Consensus and run your first search ## Using Consensus with ChatGPT Deep Research Consensus also works with ChatGPT Deep Research. When Deep Research uses the Consensus app, it can search Consensus for relevant papers and retrieve full paper details for cited results. **1. Connect the Consensus app** Connect the Consensus app in ChatGPT using the steps above. **2. Start a Deep Research request** Start a Deep Research request in ChatGPT. **3. Ask ChatGPT to use Consensus** Ask ChatGPT to use Consensus for peer-reviewed research, for example: > Use Consensus to research the evidence on exercise for depression. Prioritize systematic reviews, meta-analyses, and randomized controlled trials. Deep Research may run multiple Consensus searches, refine filters, retrieve paper details, and return a cited synthesis with direct links to papers on Consensus. ## Verify Your Connection Try one of these in your connected client. Each link opens the question pre-filled: | Ask | Try it | | - | - | | "What does the research say about creatine and cognitive function?" | [Claude](https://claude.ai/new?q=What%20does%20the%20research%20say%20about%20creatine%20and%20cognitive%20function%3F) · [ChatGPT](https://chatgpt.com/?q=What%20does%20the%20research%20say%20about%20creatine%20and%20cognitive%20function%3F) | | "Find meta-analyses on mindfulness for anxiety published since 2020" | [Claude](https://claude.ai/new?q=Find%20meta-analyses%20on%20mindfulness%20for%20anxiety%20published%20since%202020) · [ChatGPT](https://chatgpt.com/?q=Find%20meta-analyses%20on%20mindfulness%20for%20anxiety%20published%20since%202020) | | "Are there RCTs on intermittent fasting with more than 500 participants?" | [Claude](https://claude.ai/new?q=Are%20there%20RCTs%20on%20intermittent%20fasting%20with%20more%20than%20500%20participants%3F) · [ChatGPT](https://chatgpt.com/?q=Are%20there%20RCTs%20on%20intermittent%20fasting%20with%20more%20than%20500%20participants%3F) | | "What is the strongest evidence for and against remote work productivity?" | [Claude](https://claude.ai/new?q=What%20is%20the%20strongest%20evidence%20for%20and%20against%20remote%20work%20productivity%3F) · [ChatGPT](https://chatgpt.com/?q=What%20is%20the%20strongest%20evidence%20for%20and%20against%20remote%20work%20productivity%3F) | You should see results citing specific papers with titles, journals, years, citation counts, and direct links. ## Plans and Access Your Consensus plan, not your AI subscription, sets how many papers each search returns, your monthly calls, whether results include DOIs and full-text excerpts, and whether Threads can use `deep` mode. No account is needed to try it. > [card] [Plans and access](/mcp-plans-and-access) > > Limits for every plan, how calls are counted, additional usage, and API key access. ## Tools Consensus MCP exposes six tools. Your assistant picks the right one from your question. | Tool | What it does | | - | - | | [`search`](/mcp-tools/search) | Find ranked, peer-reviewed papers for a query in one call. | | [`get_full_text`](/mcp-tools/get-full-text) | Read the complete text of open-access papers from search results or Threads, as Markdown. | | [`create_thread`](/mcp-tools/create-thread) | Start a Thread: hand a research question to the Consensus research agent. | | [`get_thread`](/mcp-tools/get-thread) | Poll a Thread and read the agent's cited answer. | | [`add_to_thread`](/mcp-tools/add-to-thread) | Ask a follow-up in an existing Thread. | | [`find_threads`](/mcp-tools/find-threads) | List your Threads and find an earlier one. | `search` works without an account. The thread tools need a signed-in Consensus account or an API key. `get_full_text` needs a paid plan. > [card] [Quick start: Search & Threads](/mcp-tools/quickstart) > > When to use search, when to start a Thread, and how a Thread runs. ## Troubleshooting > [accordion] OAuth login is only supported for streamable HTTP servers (Codex) > > You added the server with `-- --transport http` instead of `--url`. Remove and re-add: > > ```bash theme={null} > codex mcp remove consensus > codex mcp add consensus --url https://mcp.consensus.app/mcp > codex mcp login consensus > ``` > [accordion] Authentication successful but server reconnection failed (Claude Code) > > Fully quit and restart Claude Code. Do not just use the reconnect option. Your credentials are saved and will work after restart. > [accordion] Tools stop working after a few hours (401 errors) > > Your OAuth token has expired. Re-authenticate: > > * **Claude Code:** Type `/mcp`, select Consensus, choose Clear authentication, re-authenticate, then restart Claude Code > * **Codex:** Run `codex mcp logout consensus` then `codex mcp login consensus` > * **Cursor:** Disconnect and reconnect the server in Settings then Tools and MCP > * **Claude Desktop:** Disconnect and reconnect the connector in Settings then Connectors > * **ChatGPT:** Disconnect and reconnect the Consensus app from ChatGPT Apps > * **Perplexity:** Disconnect and reconnect Consensus in Computer then Connectors > [accordion] Server shows failed or no tools appear > > * Verify the URL includes `/mcp` at the end > * Remove and re-add the server using the commands above > * Restart your client completely > [accordion] Consensus search has not been loaded yet (Claude Code) > > This is normal. Claude loads tool definitions on demand when many tools are present. Retry the query and it will work on the second attempt. > [accordion] ChatGPT Deep Research is not using Consensus > > Confirm the Consensus app is connected in ChatGPT Apps, then ask Deep Research to use Consensus explicitly. For example: > > > Use Consensus to find peer-reviewed papers on this question and cite the strongest evidence. > > Deep Research decides when to call connected tools, so it may not call Consensus for questions that are not research-focused. > [accordion] No results found > > Try broadening your search query or removing filters. Use specific academic terminology rather than casual language. > [accordion] Monthly call limit reached > > Paid plans can continue making calls past the included monthly limit when the account has an active metered API subscription and additional usage is enabled; extra usage is billed at \$0.05 per call, with each call covering up to 100 papers. Free plans must wait for the next monthly reset or upgrade. > [accordion] Rate limit exceeded (429 error) > > Rate limits are per user for authenticated MCP clients. A 429 that says you're sending requests too quickly is a per-minute rate limit; wait a few moments and add backoff between requests. A 429 that mentions monthly call credits is a usage limit, not a rate limit. ## Security * **Endpoint:** Only connect to the official URL: `https://mcp.consensus.app/mcp` * **API keys:** Store securely using environment variables. Never share keys in prompts or public conversations. Rotate if compromised. * **Data handling:** Search queries are sent to the Consensus API over HTTPS/TLS. Results contain publicly available academic paper metadata. No personal data is stored from MCP requests. ## Supported Platforms | Platform | Auth | Status | | - | - | - | | Claude Desktop | OAuth | Supported | | ChatGPT | OAuth | Supported | | ChatGPT Deep Research | OAuth | Supported | | Claude Code | OAuth | Supported | | Codex | OAuth | Supported | | Cursor | OAuth | Supported | | VS Code (GitHub Copilot) | Config | Supported | | Windsurf | Config | Supported | | Gemini Spark | OAuth | Supported | | Perplexity | OAuth | Supported | | Other MCP Clients | Bearer token | Supported | ## FAQ > [accordion] Is there a PubMed MCP server with semantic search? > > Yes. Consensus covers all of PubMed plus the rest of science, ranked semantically rather than by keyword match alone. You can filter by study design, sample size, journal quartile and citation count. See [Where the papers come from](/get-started/where-papers-come-from). > [accordion] What's the best MCP server for scientific literature in Claude? > > Consensus is in the [Claude connectors directory](https://claude.ai/directory/65247229-f0c7-49df-9044-fcbb8b3894c6) and works in Claude on the web, the Claude desktop app and [Claude Code](/mcp-clients/claude-code). > [accordion] How do I add a scientific search connector to ChatGPT? > > Connect Consensus from the [ChatGPT app directory](https://chatgpt.com/plugins/plugin_asdk_app_6943e6f4a928819195962de16fb9ffe4). It also works with [ChatGPT Deep Research](#using-consensus-with-chatgpt-deep-research). > [accordion] Can I use a research paper search MCP server in Cursor or VS Code? > > Yes. Install it in one click in [Cursor](/mcp-clients/cursor), or add the server URL to your MCP configuration in [VS Code (GitHub Copilot)](/mcp-clients/vs-code-github-copilot). > [accordion] Which MCP server works for pharma and drug discovery? > > Consensus is SOC 2 Type II compliant, doesn't use customer data to train AI models, and offers a zero-data-retention option on enterprise agreements. It runs behind [AI gateways](/enterprise/gateways/overview) and in [Microsoft 365 Copilot](/enterprise/microsoft-365-copilot). Owkin uses it in its agentic AI scientist. See [Security and data handling](/enterprise/security). > [accordion] Does it work for systematic reviews and meta-analyses? > > Yes. Filter by study design, including RCTs, systematic reviews and meta-analyses, and run deep literature reviews with Threads. See the [Threads quick start](/mcp-tools/quickstart). > [accordion] Is this the same as Consensus (goconsensus.com)? > > No. This is Consensus, the AI academic search engine at [consensus.app](https://consensus.app). It's unrelated to the sales demo platform at goconsensus.com. ## Additional Resources > [card] [Consensus Website](https://consensus.app) > > Explore the Consensus research platform. > [card] [API Documentation](/api-reference/search) > > Structured JSON results for pipelines and integrations. > [card] [MCP Specification](https://modelcontextprotocol.io) > > Learn more about the Model Context Protocol. > [card] [MCP server on GitHub](https://github.com/Consensus-NLP/consensus-mcp) > > View the source, report issues, and star the Consensus MCP server. # Citation Graph Source: https://docs.consensus.app/core-features/citation-graph Citation Graph is a visual way to explore how papers connect. Start with up to 5 seed papers you trust and discover related work through citation relationships instead of keywords. Citation Graph maps how research papers connect through citations. Instead of running a keyword search, you start with papers you already trust (**seeds**) and discover related work based on how the literature cites itself. Use it to find papers you missed, identify foundational and follow-up research, see how a field evolved, and spot clusters and gaps. It's included with every Consensus account. [First Look: Citation Graph Beta](https://www.youtube.com/embed/veNfFAEbTJs) ## How to build a graph **1. Open the graph** Click **Graph** in the left sidebar, or click **Citation Graph** on any paper's sidebar to seed the graph with that paper. **2. Add seeds** Add up to 5 seed papers from your recently saved papers or by searching from the Graph page. One seed traces the lineage of an idea in both directions; several seeds find the overlap between literatures. **3. Choose density** Show 20, 40, or 60 papers. **4. Explore** Hover a paper to see its citation connections. Click into a paper for details. **Promote** a result to a seed to regenerate the graph around it, or **Exclude** papers you don't want. **5. Chat with the graph** Click **Chat** at the bottom of the page. All seeds and results are attached as context, so you can ask how papers connect, how the research built on earlier work, or what patterns emerge. ## How papers are ranked Citation Graph doesn't use keyword matching. It combines three citation signals: * **Direct citations** — the paper cites your seed, or your seed cites it. * **Co-citation** — how often the paper and your seed appear together in other papers' reference lists. * **Bibliographic coupling** — how many references the paper shares with your seeds. A traditional citation crawl only looks forward or backward from one paper. Combining the three signals surfaces related work a crawl would miss. ![A citation graph on coffee brewing methods with three colored seed clusters connected by citation arrows](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/citation-graph-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=fa81f851d9a177afa1055c601ebfc3e4) Three seeds and their clusters. Older papers sit left, newer right; node size is citation count. ## Reading the graph | Element | Meaning | | - | - | | Horizontal axis | Time — older papers on the left, newer on the right | | Vertical axis | Approximate citation strength | | Node size | Citation count | | Color | Seed cluster — papers one hop from a seed take that seed's color; gray nodes bridge clusters | | Opacity | How connected the paper is to the seeds | | Arrows | Citation direction — an arrow points from a paper to the paper it cites | A legend sits in the top-right corner. You can also sort and filter by citation count, journal rank (SJR), study design, and sample size. ## Questions to ask the graph ```text theme={null} What do these papers say about [topic], and where do they disagree? ``` ```text theme={null} Which of these papers used a randomized design? ``` ```text theme={null} How did the later papers build on the earliest seed? ``` > [!NOTE] > > Citation Graph is in beta. Graphs aren't saved yet; closing or refreshing the page clears the graph. ## Related > [card] [Research Agent](/core-features/research-agent) > > Ask for citation chaining in plain language. > [card] [Library](/core-features/library) > > Save graph results into a Collection. # Check a claim Source: https://docs.consensus.app/core-features/citation-grounding Citation Grounding shows the exact quote behind every citation in a Consensus answer. Hover to see the passage, click to jump to it in the paper. Every claim in a Consensus answer is tied to a real paper. Citation Grounding goes one step further: it shows you the exact passage in that paper that supports the claim. ![A citation card over a Consensus answer, with the Full Text used badge and the supporting quote from the paper's results section highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/citation-card.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=f667ec73080c1427293b47e81dd8c534) Hover a citation to open its card. Full Text used means the quote came from the full paper, not just the abstract. It's a verification layer. Rather than asking you to trust AI search on faith, Consensus lets you check the evidence behind each sentence in seconds. ## How it works * **Hover a citation** to see the quote from the paper that supports the statement. * **Open the Evidence tab** on a paper (click its **Supporting Quotes** in References) to see every quote Consensus pulled from it for your search. * **Open the Attachment tab** to read the PDF, and pick a quote from the menu at the bottom to highlight it in its original context. Grounding works best on papers where Consensus has full text. Through [publisher partnerships](/get-started/full-text-access#where-full-text-comes-from) (Wiley, Taylor & Francis, SAGE, ACS, APA, AAAS, Oxford University Press, and others), open-access PDFs, and your own uploads, Consensus can look beyond the abstract to find the supporting passage. See [Full text access](/get-started/full-text-access). ![The paper panel on its Evidence tab, listing five numbered quotes from the paper's results section](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/paper-evidence-tab.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=c5784885d7d98f87314068e189e8c999) The Evidence tab lists every quote the answer used from this paper. ![The paper panel open on its Overview tab, with a supporting quote from the abstract highlighted and the Highlighted quote #2 picker below it](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/paper-highlighted-quote.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=54a40fe9073aef1885d478c36455ba97) Pick a quote and Consensus highlights it where it appears: in the abstract, as here, or in the PDF. ## Reading the signals | Signal | Meaning | | - | - | | Document icon with a tick next to a citation | The full text of the paper was used in the analysis | | No icon | Only the abstract was used | | Colored citations | When a Consensus Meter is shown, citations are colored by the paper's stance | The **Citations** panel in a thread lists every paper cited across the conversation. ## Where it appears Citation Grounding applies to AI-generated answers in the Consensus web app: Pro messages and Deep Reviews. The MCP server and API return paper metadata and key takeaways rather than per-claim quotes. ## Why it matters Consensus is built on a simple rule: if the system can't ground an answer in real evidence, it won't make one up. Citation Grounding makes that rule visible. You can audit any sentence, catch a paper cited for something it never claimed, and move from an answer to the primary source without leaving the page. > [!TIP] > > Combine Citation Grounding with a [citation verification workflow](/use-cases/citation-verification) to audit an entire bibliography. ## Related > [card] [Read the full paper](/get-started/full-text-access) > > How Consensus reads paywalled papers. > [card] [Research Agent](/core-features/research-agent) > > The agent that produces grounded answers. # Deep Review Source: https://docs.consensus.app/core-features/deep-search Deep is a longer, more robust version of a Consensus search. It runs up to 20 searches in parallel, uses multiple Research Agent tools like citation crawl, and reads 50 to 100 papers before answering. A Deep Review (the **Deep** button in the search bar) is an extended version of a Pro search. Where Pro analyzes up to 20 papers, Deep runs up to 20 searches in parallel, uses multiple [Research Agent](/core-features/research-agent) tools like citation crawl, includes multiple visualizations, and reads anywhere from 50 to 100 papers before it answers. Think of it as a longer, more robust version of a search, not a different product. ## How it works Deep plans its own steps. A single run can involve 30 or more: parallel searches for the core question, follow-up searches on alternate terms, citation-graph traversal from the strongest results, filters such as minimum citations or study type, and full-text reading of the papers it includes. **1. Searches in parallel** Runs up to 20 searches on your question at once, including foundational work, alternate terminology, and counterpoints. **2. Chains Research Agent tools** Follows citation connections from the initial results, applies academic filters, and runs more searches to fill gaps it finds along the way. **3. Screens the candidates** Narrows millions of retrieved papers to a few thousand eligible ones, then to the set it will read. **4. Reads and answers** Reads 50 to 100 papers, full text where available, and writes a citation-backed answer. ![The Deep step panel showing 121M retrieved, 2.6K eligible, and 50 included, with the searches and citation-graph step it ran](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/deep-search-steps.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=5339244798bc7c1366b0156fa02c9c58) A Deep run: 121M papers retrieved, 2.6K eligible, 50 included. Parallel searches, a citation-graph step seeded from 50 papers, alternate-term searches, and citation filters, across 31 steps. ## What you get The answer looks like a Pro answer, only longer and backed by more papers. Its shape follows your prompt: a synthesis, a comparison, a critique, a reading list, an evidence map, or a custom report. Answers include interactive visuals that: * Show the consensus among the research * Highlight key authors and top contributors * Extract key claims and their supporting evidence * Reveal potential research gaps Every statement carries an in-line citation with [Citation Grounding](/core-features/citation-grounding). Export with **Copy text**, **Copy with citations** (APA, MLA, Chicago, Harvard, LaTeX, AMA/Numeric), or **Export as PDF**. Tables, charts, citations, and headings survive a paste into Word, Docs, Notion, or Obsidian. ![The Export menu with Copy text, Copy with citations, and Export as PDF](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/deep-review-04.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ddb54a5a078f91f1dbff3598da7d5e1f) The export button sits under every answer. > [!NOTE] > > Deep may read more papers than it directly cites in the final answer. ## How to run one Turn on **Deep** in the search bar and the search runs as a Deep Review. That's it. You can also click it on an existing Pro message to upgrade it. ![The Consensus search bar with the Deep toggle switched on and highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/deep-toggle-on.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=66908db5a6f87b4ac22c98f390e9bd2a) Turn on Deep, then ask your question. Deep takes a few minutes. Consensus emails you when the answer is ready. Deep also runs over your [Library](/core-features/library) and Collections. Attach papers and ask for a PICO or SPIDER review, a reading list, a scoping map, or a custom report. ## When to use it | Mode | Papers analyzed | Best for | | - | - | - | | Paper search | — (list only) | Browsing sources without an AI summary | | Pro | Up to 20 | A thoughtful summary and connections between findings | | Deep | 50–100 | Complex or broad questions, validating ideas, finding gaps | Any search can run in Deep mode, but it earns its keep on questions that need many searches and many papers to answer well. Two examples: **A PICO question**, where each element deserves its own searches and the answer needs sample sizes and follow-up periods side by side: ```text theme={null} In adults with type 2 diabetes, does a Mediterranean diet, compared with a low-fat diet, improve HbA1c and cardiovascular risk markers? Focus on RCTs and meta-analyses from the last 10 years. Report sample size and follow-up duration for each trial, and flag where the evidence conflicts. ``` **Starting from a paper**, where Deep crawls the citation graph in both directions and organizes what it finds over time: ```text theme={null} Start from Hanahan and Weinberg (2000), "The Hallmarks of Cancer." Trace the papers it built on and the key papers that cite it, then break down how the field has evolved over the past 20 years: the major breakthroughs by year, which hallmarks were added or revised, and where the strongest current evidence points. ``` ## Availability | Plan | Deep Reviews per month | | - | - | | Free | 3 | | Pro | 15 | | Deep | 200 | | Teams | 50 per user | | Enterprise | Custom | > [!NOTE] > > On the Free plan, a Deep Review also uses one of your 10 monthly Pro message credits, because Deep is an upgraded Pro message. Check your usage under **Settings › Subscription**. You can also run a Deep Review from your AI assistant or code: the [MCP server](/mcp-tools/quickstart) and [API](/api-quickstart-threads) run Research Agent as a Thread, and `deep` mode runs the longer review on paid plans. ChatGPT's "Deep Research" is a ChatGPT feature that can call Consensus search; it is not the same as Deep in Consensus. ## Related > [card] [Research Agent](/core-features/research-agent) > > The engine behind Pro and Deep. > [card] [Library](/core-features/library) > > Run a Deep Review over your own papers. # Extract data Source: https://docs.consensus.app/core-features/extract-data Extract data turns a set of papers into a table that Consensus fills in for you. Each paper is a row, each column is a question, and every answer links to the quote it came from. [YouTube video player](https://www.youtube.com/embed/SisukwyizMI) Extract data turns a set of papers into a table that Consensus fills in for you. Each paper is a row and each column is a question, like sample size or funding source. Consensus reads the papers, answers each cell, and shows you the quote behind every answer. Use it to compare studies, pull the details you need for a literature review, or check what a set of papers reports, without opening every PDF yourself. Extract data is included on every plan. ## Start a table There are three ways to create a table: * **From search results.** Select the papers you want, then click **Extract data** at the bottom of the results. * **From My Library or a Collection.** Select papers, then click **Extract data**. * **From the sidebar.** Click **Extract data** under Tools, choose a blank table or a template, then add papers from My Library, upload files, or paste DOIs. A table can hold up to 5,000 papers. ### Templates Templates start your table with a set of columns already added. | Template | Columns | Use it when | | - | - | - | | **Quick summary** | Key takeaway, Year, Journal, Authors | You want an overview of a set of papers | | **PICO** | Population, Intervention, Comparison, Outcome measured | You're comparing clinical or intervention studies | | **Study characteristics** | Study Design, Sample size, Population, Outcome measured | You're describing the studies in a review | | **Critical appraisal** | Study Design, Sample size, Limitations, Funding or conflicts of interest | You're assessing the strength of each study | ## Add papers Click **Add papers** at the top of the table and choose: > [accordion] From My Library > > Pick saved papers or whole Collections. > [accordion] Upload > > Upload PDF, RIS, or BibTeX files. Anything you upload is also saved to My Library. If you upload the PDF of a paper that's already in the table, it's added to that paper's row rather than creating a new one. > [accordion] Paste DOIs > > Paste one or more DOIs. These papers are added to the table only, not to My Library. ## Add columns Click **+ Column** in the table header to add a suggested column or create your own. Extracted columns are answered by reading each paper, and use credits: * Sample size * Population * Intervention * Outcome measured * Limitations * Funding or conflicts of interest Metadata columns come from the Consensus index rather than the paper itself. They're free and fill in right away: * Authors * Year * Journal * Citation count * Full text? * Key takeaway * Study Design Papers that aren't in the Consensus index, such as some uploads and dissertations, may have blank metadata columns. ### Create a custom column **1. Open the column menu** Click **+ Column**, then **Create custom column**. **2. Name it and choose a format** Pick **Free text**, **Number**, **Yes or No**, **Single tag**, or **Multiple tags**. For tag formats, list at least two possible values for Consensus to choose from. **3. Write your question** Type it under **Instructions**, the way you'd ask a colleague. **4. Choose whether to require full text** Turn on **Require full text** if the answer is unlikely to be in the abstract, such as details from the methods section. Otherwise, Consensus checks the abstract first. Custom columns people create: ```text theme={null} theme={null} How many participants were enrolled? ``` ```text theme={null} theme={null} Was the study preregistered? ``` ```text theme={null} theme={null} Which outcome measure was used for depression symptoms? ``` ```text theme={null} theme={null} What future areas of study do the authors suggest? ``` > [!TIP] > > Ask one question per column. Separate **Sample size** and **Mean age** columns will give you better answers than one column asking for both. ## How cells are filled Consensus reads each paper's abstract first. If the abstract doesn't answer the question, it reads the full text. Columns with **Require full text** turned on, along with **Limitations** and **Funding or conflicts of interest**, go straight to the full text. Every answer comes from a sentence in the paper about its own study, not from the other studies it cites. If the paper doesn't report something, the cell says so. While the table fills, the header shows **Running M / N cells**. When it's done, a banner shows how many cells were filled. ### What each cell state means | State | Meaning | | - | - | | **Found in abstract** / **Found in full text** | Answered, and where the answer came from | | **Not found in abstract** / **Not found in full text** | Consensus read the paper and it doesn't report this | | **Needs full text** | The answer is likely in the full text, but Consensus can't access it | | **Not run** | The cell hasn't been run yet, usually because you've run out of credits | | **Failed** | Something went wrong. Click **Run pending** to try again | | **Needs credits** | You don't have enough credits to run this cell | ## Check an answer Click any cell to see what Consensus read (the abstract, the full text, or both) and what it found. Click **View evidence** to see the supporting quotes in the paper. ## Add the full text yourself Hover over a **Needs full text** cell. If a PDF is available through your library or open access, you'll see a link to it. If you have the PDF, click **Upload full text**. Consensus re-reads every **Needs full text** cell in that row for free, and saves the PDF to My Library. Only the person who owns the table can upload full text. ## Filter and organize the table Use the filters above the table to show only rows that are **Finished**, **Needs full text**, **Not run**, or **Failed**. A row can appear under more than one filter. To find a specific paper, search the table. Open a column's menu to see how many of its cells are finished, running, or waiting, and to view, hide, or delete the column. Drag columns to reorder them. ## Cells and credits Each extracted cell uses one credit, whether Consensus needs only the abstract or the full text too. Metadata columns are free, and so are re-reads after you upload full text. | Plan | Extracted cells per month | | - | - | | Free | 100 | | Pro | 500 | | Teams | 500 | | Enterprise | 500 | | Deep | 10,000 | **Run pending ⚡N**, in the table header or a column's menu, runs every cell that hasn't been answered yet or failed. N is the number of credits it'll use; hover over it to see the breakdown. If you don't have enough credits for every cell, Consensus fills as many as it can. Your credit balance is shown next to **Share**. ## Save, share, and export * **Recents** — every table you create appears in **Recents** in the sidebar. * **Save** — add a table to My Library or a Collection. * **Export** — download the table as a CSV. * **Share** — click **Share** and set the table to **Public** so anyone with the link can view it, or keep it **Private**. * **Make a copy** — anyone who can view a table can copy it and keep working on their own version. Copying doesn't re-run any cells, so it uses no credits. ## Tables in a Deep Review If you ask for a table during a [Deep Review](/core-features/deep-search), Research Agent can build an extraction table as part of its answer. The table appears as a card under the answer, and the answer cites its cells. These tables are read-only and don't appear in Recents. To edit one, open it and click **Make a copy**. ## Limits * Each cell holds one answer per paper, so papers that report several studies or samples may not show all of them. * Consensus only answers from text it can quote. Questions that depend on interpreting a whole paper, which is common with qualitative or theoretical work, may come back empty or with the closest related quote. * Text answers are capped at about 2,000 characters, and columns asking Consensus to copy a paper's text word for word aren't allowed. ## Related > [card] [Library](/core-features/library) > > Save papers and Collections to build tables from. > [card] [Deep Review](/core-features/deep-search) > > Ask for a table and let Research Agent build it. # Library Source: https://docs.consensus.app/core-features/library My Library is where you save, organize, and work with research in Consensus. Store papers, searches, and uploaded documents in Collections, import from Zotero, and chat with everything you've collected. My Library is a reference manager that goes beyond storage. Save papers and searches, upload your own PDFs, import your Zotero library, then search and reason across everything you've collected. Library is included on every plan. It's private by default. ![My Library showing saved papers in a table and the Ask papers in My Library search bar](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/my-library-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=e7ec69dfe4bc32584db66671d90a3cf5) ## Organize with Collections Save items to **My Favorites** or create custom Collections by topic, project, or question. Collections can be nested: when you create one, pick a parent under **Create in**. Inside a Collection, sort by Title, Authors, Journal, Year, DOI, Design, or Saved date. ![My Library with the Collections list in the left sidebar and the New Collection button highlighted](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/my-library-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=46213478fef550ff76bc60b62ce45f65) Collections live in the left sidebar; the folder-plus button creates a new one. ## Add papers Click **+ Add** in the top-right corner of My Library and choose: ![My Library with the + Add menu open, showing Upload, Paste DOIs, and Import from Zotero](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/library-add-menu.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=5318d155b8f455ddcd2c5cf49558ca8c) > [accordion] Upload files > > Upload PDF, RIS, or BibTeX files, up to 20 at a time. Published papers work, and so do your own documents: manuscripts, proposals, dissertations, clinical protocols, and private reports. > > Uploads must be text-based PDFs under 50 MB. Scanned or image-only PDFs can't be processed. > [accordion] Paste DOIs > > Paste one or more DOIs to add the matching papers. > [accordion] Import from Zotero > > Click **Import Zotero**, create an API key on Zotero (the form is pre-filled), paste the key, and import. Both personal and group libraries are supported, with no size limit. > > ![My Library sidebar showing a connected Zotero collection next to a regular collection](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/zotero-import-05.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=f0db3b6d3587c9218cb85a9891149049) > > Connected collections (blue) keep your Zotero structure; regular collections are black. > > Imported collections appear as blue **Connected collections** that preserve your Zotero folder structure. Import is one-way and manual: re-import to pull in anything added since your last import. Only journal articles are used in search and analysis, and only PDFs stored in Zotero (not externally linked files) are imported. > [accordion] Save from search > > Open a paper (click its **Supporting Quotes** in References, or **Ask** on a citation) and click **Save**. The paper is added to **My Library** straight away. Tick a Collection in the same menu to file it there too, or type in the search box to find one. To save a whole search, click the bookmark at the top of the thread. > > ![The Save menu with My Library ticked and a search box for adding the paper to a Collection](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/save-menu.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=e76aae5a5353d5737f0e35be0fdce8aa) ## Search and chat with your Library There are two ways in: * **From a new search**, click **+** › **Search corpus** › **My Library** to search everything you've saved, or **+** › **Attach papers or collections** to pick a specific Collection or set of papers. * **From inside My Library**, ask a question in the search bar at the bottom of the page. It says **Ask these papers…** and is already scoped to your Library; inside a Collection, it uses only that Collection. ![The My Library search bar with My Library attached, the Ask these papers prompt, and the Corpus button highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/library-search-bar.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=fe9b1cd03793da006127418ce6b66760) Inside My Library, the search bar is already scoped to your saved papers. Both Pro and Deep modes work on your Library. A Deep Review over a Collection analyzes up to 50 papers and can produce PICO or SPIDER reviews, reading lists, scoping maps, and custom reports. Things people ask their Library: ```text theme={null} Which papers in this collection support the claim that [statement]? ``` ```text theme={null} Build a table showing which of these studies support or challenge [hypothesis]. ``` ```text theme={null} What populations am I missing in this collection? ``` > [!NOTE] > > Chatting with a Collection or your Library runs as a Pro or Deep message and uses those credits. See [Research Agent](/core-features/research-agent) for plan limits. ## Manage your Library from the API or MCP The research agent can also work in your Library when you call it through the [Consensus API](/api-quickstart-threads#work-in-your-library) or [MCP](/mcp-tools/create-thread#library-actions). Ask it to list your Collections and their papers, create or rename a Collection, or save the papers it found. Removing papers, deleting a Collection, and moving one stay in the web app. ## Share a Collection Click **Share** on a Collection to invite people by email or copy an invite link. **Viewers** can read the papers and run searches on the Collection; **Editors** can also rename it, add sub-Collections, and add or remove papers. Anyone without an account is prompted to create a free one. Shared Collections appear under **Shared with me**. Saved threads inside a Collection aren't shared. ![A Collection open in My Library, with the Share button in the top right highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/share-collection-01.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ab61c1adbfdbc0383aa1cc63d73ce496) ## Export * **Bibliography** — select papers, click the download button, and choose APA, MLA, Chicago, Harvard, LaTeX, or AMA/Numeric. * **CSV or RIS** — export a Collection or your whole Library for EndNote, Mendeley, Zotero, or Paperpile. ![Several papers selected in My Library with the export menu showing CSV, RIS, and Bibliography](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/my-library-08.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=2645ddb12e91325c4508afb7af7a51ce) > [!NOTE] > > The [MCP server](/core-features/mcp) and [API](/api-get-started) search the Consensus research index, not your Library. ## Related > [card] [Deep Review](/core-features/deep-search) > > Run a full review over a Collection. > [card] [Citation Graph](/core-features/citation-graph) > > Seed a graph from your saved papers. # Use Consensus in ChatGPT, Claude, Copilot, or with the API Source: https://docs.consensus.app/core-features/mcp Connect Consensus to ChatGPT, Claude, or Microsoft 365 Copilot in one click, or build on it with the API, so answers cite real peer-reviewed papers. If you already use ChatGPT, Claude, or Microsoft 365 Copilot, connect Consensus to it. Your assistant can then search over 400 million scholarly articles, including 350 million peer-reviewed research papers, apply academic filters, hand research questions to the Consensus research agent, and cite real research instead of answering from memory. ## Connect in one click Pick your assistant and add Consensus from its app directory. > [card] [ChatGPT](https://chatgpt.com/plugins/plugin_asdk_app_6943e6f4a928819195962de16fb9ffe4) > > Add Consensus from the ChatGPT app directory. > [card] [Claude](https://claude.ai/directory/65247229-f0c7-49df-9044-fcbb8b3894c6) > > Add Consensus from the Claude connectors directory. > [card] [Microsoft 365 Copilot](https://marketplace.microsoft.com/en-us/product/WA200011861) > > Add Consensus for Copilot from the Microsoft Marketplace. Sign in with your Consensus account the first time you connect. Your results depend on your Consensus plan, not your ChatGPT, Claude, or Copilot subscription. ### Connect another tool Using a different assistant or tool, such as Cursor, VS Code, or Claude Code? Connect it to the Consensus MCP server directly. MCP (Model Context Protocol) is a standard way for AI assistants to use outside tools. The server is hosted, so there's nothing to install: ```text theme={null} https://mcp.consensus.app/mcp ``` > [card] [Set up your client](/consensus-mcp) > > Step-by-step connection guides for Claude Code, Cursor, VS Code, Windsurf, Gemini Spark, Copilot Studio, and any other MCP client. ## What your assistant can do **Search.** The `search` tool returns ranked papers with title, authors, journal, publication year, citation count, study type, a key takeaway, and a link to the paper on Consensus. On paid plans it also returns query-relevant excerpts from open-access papers, so your assistant can quote the passage behind each claim. Your assistant can pass the same filters Research Agent uses: * Year range, study type, and journal quartile (SJR) * Human and controlled studies, minimum sample size, minimum citations * Medical mode, open access only, exclude preprints * Study duration and country **Read whole papers.** On paid plans, `get_full_text` reads the complete text of open-access papers from search results or Threads, so your assistant can work from a paper's methods and results, not just its abstract. **Threads.** With a signed-in account, your assistant can also run [Research Agent](/core-features/research-agent) as a Thread. `create_thread` starts the agent on a research question, `get_thread` returns the cited answer when it's ready, `add_to_thread` asks a follow-up, and `find_threads` finds earlier Threads. The agent can search your Library collections and save papers to them, and `deep` mode runs a longer literature review on paid plans. > [card] [Quick start: Search & Threads](/mcp-tools/quickstart) > > When to use each, with a reference page for every tool. ## Supported clients Claude (web, desktop, and Claude Code), ChatGPT and ChatGPT Deep Research, Codex, Cursor, VS Code and GitHub Copilot, Windsurf, Gemini Spark, Microsoft 365 Copilot and Copilot Studio, and any client that supports remote MCP servers. > [tab] Claude > > Open **Settings › Connectors**, search for Consensus, and click **+**. Then switch it on from the **+** menu in any chat under **Connectors**. > > ![The Claude Connectors dialog with Consensus found by search and the add button highlighted](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/mcp-claude-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=4af6e73763cbcfeba2cadff6f81e741c) > > ![The Claude chat + menu with the Consensus connector toggled on](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/mcp-claude-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=f27608d3d3497d44f64aae30dd1b011b) > [tab] ChatGPT > > Open **Apps**, search for Consensus, and connect. It also works inside **Deep research**: open **Sources** and toggle Consensus on. > > ![The ChatGPT Apps page with Consensus found by search](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/mcp-chatgpt-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=6524b1e5113f30a3b50d2f0b2d634200) > > ![The ChatGPT Deep research composer with the Sources menu open and Consensus toggled on](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/mcp-chatgpt-03.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=1c1a17e5bf71b5f74af7356e8c998f1c) > > In ChatGPT Deep research, Consensus appears under Sources. > [tab] Microsoft 365 Copilot > > In Microsoft 365 Copilot, Consensus is a connector your admin adds for your organization. Once it's there: > > 1. Select the **gear icon** at the bottom of the left sidebar, then **Settings › Sources**. > 2. Find **Consensus for Copilot**, select **Connect**, and sign in with your Consensus account. > 3. Ask a research question, and Copilot calls Consensus to find relevant papers. > > Don't see Consensus under **Sources**? Send your admin the [Microsoft 365 Copilot admin guide](/enterprise/microsoft-365-copilot). You can use the MCP server without an account at reduced limits. ## Pre-built workflows Consensus publishes prompt templates ("skills") for common research jobs you can run through MCP: * **Literature review helper** — reconnaissance, PICO/SPIDER decomposition, systematic searching, and synthesis. * **Curriculum development** — parse a syllabus into a reading list with discussion questions. * **Grant research** — position your idea, map it to NIH institutes and mechanisms, and find funded overlap. See [Skills](/core-features/skills), or browse the full gallery in [Use cases](/use-cases). ## Use the API Building your own tool, agent, or workflow? The Consensus API gives you the same paper search over REST, and can run Research Agent as a Thread. Send a query, get back ranked peer-reviewed papers with metadata, and skip running each search by hand. It's self-serve and open to everyone. ```text theme={null} GET https://api.consensus.app/v1/search ``` * The top 20 most relevant papers per query by default; raise `page_size` for more, up to your plan's maximum * Metadata on every paper: title, authors, journal, year, DOI, citation count, and a link * On paid plans: study design, journal rank, key takeaways, and query-relevant excerpts from open-access papers * The same filters as the web app: publication date, study type, sample size, citations, journal quality, field, country, and open access To get a key, click your profile icon in Consensus, choose **API & MCP Dashboard**, open **Keys and Clients**, and click **New key**. > [card] [Get started with the API](/api-get-started) > > Authentication, your first request, and the full parameter reference. > [card] [Run Research Agent from code](/api-quickstart-threads) > > Start a Thread, wait for the cited answer, and ask follow-ups. ## Plans and limits API and MCP calls share one monthly pool. | Plan | Max papers per call | Included calls per month | Rate limit | Extra fields | | - | - | - | - | - | | No account | 3 | No monthly limit | 5 per second, shared | — | | Free | 10 | 30 | 1 per second | — | | Pro | 300 | 500 | 3 per second | Study type, takeaway, DOI | | Deep | 750 | 2,000 | 3 per second | Study type, takeaway, DOI | | Teams | 300 | 500 | 3 per second | Study type, takeaway, DOI | | Enterprise | 1,000 | Custom | Custom | Study type, takeaway, DOI | * A call covers up to 100 papers: 1–100 papers is 1 call, 101–200 is 2, and so on. Requests default to 20 papers. * Paid plans can enable **additional usage** at \$0.05 per call, up to 1,000 calls per month (10,000 on Deep). It's off by default. * Usage resets on the 1st of each month. See [MCP plans and access](/mcp-plans-and-access) for the full breakdown. Track usage, connected clients, and billing from your profile icon under **API & MCP Dashboard**. ![The API & MCP Dashboard's Keys and clients tab listing Claude and ChatGPT as connected MCP clients](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/api-03.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=099f4dd0120f71a4728a35d5b3c767ef) Connected MCP clients appear under Keys and clients, alongside your API keys. > [!NOTE] > > Clients without OAuth can pass any Consensus API key as a Bearer token. The key gets its owner's plan limits. See [MCP plans and access](/mcp-plans-and-access#api-key-access). For custom Enterprise limits, contact [sales@consensus.app](mailto:sales@consensus.app). ## Security Only connect to the official URL above. Queries travel over HTTPS, results contain publicly available paper metadata, and no personal data is stored from MCP requests. ## Related > [card] [Get started with MCP](/consensus-mcp) > > Connect your client in under 2 minutes. > [card] [API](/api-get-started) > > The same search over REST. # Research Agent Source: https://docs.consensus.app/core-features/research-agent Research Agent is the AI research assistant behind every Pro message and Deep Review in Consensus. It plans searches, chains tools, applies academic filters, and returns citation-backed answers. Research Agent is the engine that powers searches in Consensus. Instead of running one search at a time, it reasons across multiple tools, chains them together, and returns citation-backed answers grounded in peer-reviewed research. You don't switch it on. Every Pro message and Deep Review runs through Research Agent. ## What it does * **Plans multi-step searches.** Breaks a complex question into sub-questions and runs targeted sub-searches when a single query isn't enough. * **Applies filters in plain language.** Stack recency, journal quartile, citation count, sample size, study type, and population by describing what you want. Research Agent picks the right filters automatically. * **Traces authors and citations.** Give it a DOI, a paper title, or an author name to explore co-authors, influential papers, and how the work developed over time. * **Excludes what you don't want.** Filter out the wrong population, methodology, publication type, or author. * **Carries context between searches.** Follow-up questions build on the previous search, so you can refine, pivot, and synthesize without starting over. * **Reads and compares papers.** Reads full text where available, surfaces relevant findings, and builds comparison tables across studies. * **Finds gaps.** Detects missing populations or research areas in a Library collection or set of papers. ## Effort levels You choose how much work Research Agent does with the **Deep** toggle in the search bar. Every level searches the same 400M+ papers; higher levels run more searches and read more papers before answering. | Level | What it does | Papers read | Best for | | - | - | - | - | | Paper search | A ranked list of papers, no AI summary | — | Browsing sources yourself | | **Pro** (default, Deep off) | Plans a few searches and writes a cited answer | Up to 20 | Most questions | | **Deep** (Deep on) | A [Deep Review](/core-features/deep-search): up to 20 parallel searches, citation crawling, and a longer report | 50–100 | Broad questions, overviews, finding gaps | ## How to use it **1. Open Consensus** Go to [consensus.app](https://consensus.app). **2. Ask your question** Type a question, topic, messy prompt, or Boolean-style query in the search bar. Research Agent handles all three. **3. Send the message** By default the message runs as a Pro message. Click **Deep** to upgrade it to a Deep Review for a longer, more comprehensive analysis. **4. Follow up** Ask a follow-up in the same thread. Research Agent keeps the context, so "now only Q1 journals" or "exclude animal studies" just works. ![A Research Agent trace showing two parallel searches with year and journal filters, and retrieved, eligible, and included counts](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/research-agent-01.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=b6eef8e56acdc999c4b91cda8a2ca6e8) Every answer opens with the agent's trace: papers retrieved, eligible, and included, plus each sub-search and its filters. ## Example prompts ```text theme={null} Find studies on GLP-1 agonists and cardiovascular outcomes published in the last 3 years, in Q1 journals, with sample sizes over 500. Summarize the strongest evidence. ``` ```text theme={null} Find RCTs on CBT, IPT, and digital interventions for adolescent depression. Build a comparison table with columns: intervention type, sample size, session count, follow-up duration, and effect size on depression symptoms. ``` ```text theme={null} Starting from Jennifer Doudna's foundational paper on CRISPR-Cas9, identify the most important papers it cites that contributed to it as a genome-editing tool, then identify the key later papers that cite it. Organize results by year and type of contribution. ``` ## What you get back ![A citation card over a Consensus answer, with the Full Text used badge and the supporting quote highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/citation-card.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=f667ec73080c1427293b47e81dd8c534) Hover a citation to see its card. Full Text used means the paper's full text was analyzed. Answers appear in the thread with in-line citations that link directly to the papers supporting each statement. A document icon with a tick next to a citation means the full text of the paper was used in the analysis; no icon means the abstract was used. Depending on the question, the answer may include: * **Consensus Meter** — a "yes or no" visualization showing where the research leans. * **Claims & Evidence table** — strength of support, reasoning, and key papers for each claim. * **Research Gaps heat map** — areas with limited research. * **Results timeline** and a **TL;DR** header. Export any answer with **Copy text**, **Copy with citations** (APA, MLA, Chicago, Harvard, LaTeX, AMA/Numeric), or **Export as PDF**. The papers themselves export as CSV, RIS, or a formatted bibliography from the References panel. ![The export menu on the References panel with CSV, RIS, and Bibliography options](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/pro-messages-05.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=34fefbf55ef6088db4d48d2e849557c6) ## Availability Research Agent runs in both Pro and Deep modes. | Plan | Pro messages | Deep Reviews | | - | - | - | | Free | 10 per month | 3 per month | | Pro | Unlimited | 15 per month | | Deep | Unlimited | 200 per month | | Teams | Unlimited | 50 per month per user | > [!NOTE] > > You can also run Research Agent from your AI assistant or code. The [MCP server](/mcp-tools/quickstart) and [API](/api-quickstart-threads) run it as a Thread, alongside the underlying paper search. ## Under the hood Research Agent is a multi-agent system. A planning agent breaks down your question and decides which actions to take; a search agent combs the paper index, your private Library, and the citation graph; a reading agent interprets papers individually or in batches; and an analysis agent synthesizes the results and composes the answer. If no relevant studies meet its quality threshold, it says so rather than inventing an answer. > [!TIP] > > Research Agent replaced Scholar Agent in May 2026. If you see "Scholar Agent" in older material, it refers to the previous version of this feature. ## Related > [card] [Deep Review](/core-features/deep-search) > > The longer, more robust version of a search: 50 to 100 papers. > [card] [Check a claim against the paper](/core-features/citation-grounding) > > See the exact quote behind every citation. # Skills Source: https://docs.consensus.app/core-features/skills Skills are ready-made research workflows built into the Consensus search bar. Pick one, such as Literature Review or Verify Citations, and Consensus runs the whole task for you as a Deep Review. A skill is a ready-made workflow for one research task. Instead of writing a long, careful prompt yourself, you pick the skill and type your topic. Consensus then follows the same structured steps every time. ## Run a skill **1. Open Skills** Click **Skills** in the search bar. ![The Skills dropdown listing Literature Review, Citation Snowballing, Reverse Citation Search, and Verify Citations](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/skills-menu.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=a0bcb3df0ee400dd07d5be48fc29ac47) **2. Pick a skill** Choose one of the skills below. The **Skills** button changes to show the skill's name, and **Deep** turns on automatically, because every skill runs as a [Deep Review](/core-features/deep-search). **3. Type your topic, or try an example** Each skill shows a few **TRY** prompts under the search bar. Click one to see how the skill works, or type your own topic and press Enter. ![The search bar after picking Literature Review: the skill and Deep are both highlighted, with three TRY example prompts below](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/skills-selected.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=05039f919823100d88d00e07cb3f033a) > [!NOTE] > > Because skills run as Deep Reviews, each one uses a Deep Review from your plan: 3 per month on Free, 15 on Pro, and 200 on Deep. ## Skills in the search bar | Skill | What it does | Use it when | | - | - | - | | **Literature Review** | Builds a structured search strategy and writes a report on what the research says | You're starting a new topic and need the lay of the land | | **Citation Snowballing** | Traces citations outward from seed papers you choose | You have one or two key papers and want everything connected to them | | **Reverse Citation Search** | Takes a statement you paste and finds papers that support or challenge it | You've written a claim and need sources for it | | **Verify Citations** | Checks whether cited papers actually support the claims they're attached to | You want to audit a draft, or someone else's reference list | > [!TIP] > > New to a topic? **Literature Review** is the place to start. Run it on your topic after [Run your first search](/get-started/run-your-first-search), then use the report as a map while you work through the rest of this guide. ## Skills for AI assistants If you use Consensus inside an AI assistant such as Claude or ChatGPT (see [Use Consensus in ChatGPT, Claude or Copilot](/core-features/mcp)), Consensus also publishes skills you can run there: > [card] [Literature review helper](/use-cases/literature-review-skill) > > Surveys an unfamiliar field, breaks your question down with a framework like PICO, searches systematically, and writes a research guide. > [card] [Syllabus reading list](/use-cases/syllabus-reading-list) > > Turns a course syllabus into a reading list of recent papers, with a plain-language summary and a discussion question for each. > [card] [Grant finder](/use-cases/grant-finder) > > Positions a research idea, maps it to the right NIH institutes and funding mechanisms, and finds funded projects that overlap with it. # Connection reference Source: https://docs.consensus.app/enterprise/connection-reference The technical facts every integration needs: the MCP endpoint, transport behavior, both authentication options, OAuth discovery documents, and the domains to allowlist. Everything on this page applies to any MCP client or gateway. The platform guides link back here rather than repeating it. ## Endpoint and transport | | | | - | - | | **MCP endpoint** | `https://mcp.consensus.app/mcp` | | **Transport** | Streamable HTTP | | **Responses** | Buffered JSON. There is no SSE endpoint and the server never returns SSE event streams. | | **Sessions** | Stateless. No `Mcp-Session-Id`, no session affinity, so your gateway needs no sticky-session configuration. | | **Tools** | `search`, plus the thread tools `create_thread`, `get_thread`, `add_to_thread`, and `find_threads` for signed-in accounts and API keys, and `get_full_text` for paid plans. See the [tool reference](/mcp-tools/quickstart). | > [!WARNING] > > All MCP and JSON-RPC calls must target the `/mcp` path. The bare host `https://mcp.consensus.app/` only answers health checks and OAuth discovery. Where a product lets you choose a transport, pick **Streamable HTTP** explicitly. Some gateways default to SSE, which will fail against this server. ## Authentication Two options. For a shared, server-side gateway credential, use an API key. > [tab] API key (recommended for gateways) > > A single opaque credential sent as a bearer token: > > ```http theme={null} > Authorization: Bearer ak_live_... > ``` > > It lives entirely server-side, involves no user redirect and no token refresh, and maps to your Enterprise tier. Generate one from your Consensus account: profile icon → **API & MCP Dashboard** → **Keys and clients** → **New key**. The full key is shown once. > > Treat it like a password: keep it out of client-side code, public repositories, and screenshots. > [tab] OAuth 2.0 (per-user) > > Authorization-code flow only. There is no client-credentials grant, so OAuth always involves a per-user consent step; each user signs in to their own Consensus account. > > | | | > | - | - | > | **Authorization URL** | `https://consensus.app/oauth/authorize/` | > | **Token URL** | `https://consensus.app/oauth/token/` | > | **Dynamic Client Registration** | `https://consensus.app/oauth/register/` (RFC 7591) | > | **Scope** | `search`. Request `scope=search` or omit it; it's the default. This is not an OIDC provider, so there are no `openid` or `email` scopes. | > | **PKCE** | Required. `S256` only; requests without a `code_challenge` are rejected. | > > Keep the trailing slashes on the URLs. > > **Dynamic Client Registration issues public clients**: a `client_id` with no `client_secret`, which is why PKCE is required. If your gateway needs a confidential client (a `client_id` and `client_secret` pair) and can't do DCR, Consensus provisions those manually for enterprise partners. [Contact sales](https://consensus.app/home/contact/sales/) and we'll set one up. ## Discovery documents Gateways that support automatic configuration can read these: * `https://mcp.consensus.app/.well-known/oauth-protected-resource` — points to the authorization server * `https://consensus.app/.well-known/oauth-authorization-server` — full metadata: endpoints, scopes, and `S256` ## Firewall allowlist If your organization runs an egress firewall, allow these hosts. Per-host wildcards cover any path changes on our end (new discovery documents, versioned endpoints) without another change request, while staying scoped to the hosts this integration touches: ```text theme={null} mcp.consensus.app/* consensus.app/* clerk.consensus.app/* ``` `clerk.consensus.app` is our authentication provider. Sign-in does a browser redirect through it, so it needs to be reachable wherever your users complete the consent screen. If you use the API-key path only, users never sign in and this host isn't needed. > [accordion] Tighter scope: exact URLs > > If policy requires exact URLs rather than wildcards, allow these. Expect a new change request any time we add or change a path. > > ```text theme={null} > https://mcp.consensus.app/mcp MCP JSON-RPC endpoint > https://mcp.consensus.app/ health check > https://mcp.consensus.app/.well-known/oauth-protected-resource > https://consensus.app/.well-known/oauth-authorization-server > https://consensus.app/oauth/register/ > https://consensus.app/oauth/authorize/ > https://consensus.app/oauth/token/ > https://clerk.consensus.app/* sign-in redirects > ``` ## Protocol version The server tracks the current MCP specification. Some gateways refuse upstream servers below a minimum protocol version (Kong and Azure API Management require `2025-06-18` or later). If your gateway reports a version mismatch, [contact us](https://consensus.app/home/contact/sales/) with the version it requires. ## What Enterprise unlocks | | Paid plans | Enterprise | | - | - | - | | Rate limit | 3 requests per second | Custom | | Papers per call | Up to 300 (Pro, Teams) or 750 (Deep) | Up to 1,000 | | Monthly calls | 500 to 2,000 included, with paid overage | Unlimited search; complete text and Threads billed per credit | | Full-text excerpts | Yes | Yes | # Microsoft Copilot Studio Source: https://docs.consensus.app/enterprise/copilot-studio Add the Consensus MCP server as a Model Context Protocol tool in Copilot Studio, with per-user OAuth or a shared API key, so your own agents can search 400 million papers. Copilot Studio connects to remote MCP servers directly, so you can give any agent you build the Consensus `search` tool. The tool can be shared with other makers in your environment. Copilot Studio supports the Streamable HTTP transport only, which is what Consensus uses. ## Add the tool **1. Open Tools** Open your agent in [Copilot Studio](https://copilotstudio.microsoft.com) and go to the **Tools** page. ![The Tools tab in the Copilot Studio left navigation](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.17.16@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=b5f6a6010734943bb41b7d8351b2e952) **2. Add a new tool** Select **Add a tool**, then **New tool**. ![The Add a tool dialog with New tool highlighted](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.19.50@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=cf3823beb65f9b462ad5c1f658e9b6ba) **3. Choose Model Context Protocol** ![Tool type options with Model Context Protocol highlighted](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.21.06@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=7bb99d2343cd51fa74f3ace9e886aabe) **4. Fill in the server details** ![The MCP server form with name, description, URL, and authentication fields](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.27.17@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=557f2599cb9066c2bb71177b2bc4b402) | Field | Value | | - | - | | **Server name** | Consensus | | **Server description** | Consensus is the go-to MCP for academic research. Search 400M+ scholarly articles, including 350M peer-reviewed papers, from PubMed, Semantic Scholar, and ArXiv — filter by study type (RCTs, meta-analyses, systematic reviews), journal quality, sample size, and more. Get results with citations, study takeaways, and direct links. | | **Server URL** | `https://mcp.consensus.app/mcp` | | **Authentication** | See the two options below | **5. Create, then add to the agent** Select **Create**. On **Add tool**, choose **Create a new connection** (or reuse one), then **Add to agent**. ## Choose an authentication option > [tab] OAuth 2.0 — each user signs in > > Set **Authentication** to **OAuth 2.0** and **Type** to **Dynamic discovery**. Copilot Studio reads our discovery document, registers itself with Dynamic Client Registration, and prompts each user to sign in to their own Consensus account the first time they use the tool. Usage and limits follow each user's Consensus plan. > > Copilot Studio's registration call comes from Power Platform infrastructure, so `https://consensus.app/oauth/register/` must be reachable from there (it is, unless your network policy blocks it). > [tab] API key — one shared credential > > Set **Authentication** to **API key**, **Type** to **Header**, and the header name to `Authorization`. When you create the connection, enter the value as `Bearer ` followed by your key: > > ```text theme={null} > Bearer ak_live_... > ``` > > Every user of the agent then searches through the same Consensus account and its plan limits. Generate the key from your Consensus profile icon → **API & MCP Dashboard** → **Keys and clients** → **New key**. > [!NOTE] > > If your organization requires a client ID and client secret (the **Manual** OAuth type), Consensus provisions confidential clients for enterprise partners. [Contact sales](https://consensus.app/home/contact/sales/). Don't use Manual with a Dynamic-Client-Registration client: token refresh will fail and calls will start returning 401 on the second turn. ## Do I have Copilot Studio access? Go to [copilotstudio.microsoft.com](https://copilotstudio.microsoft.com) and sign in with your work account. If the **Home** page loads, you're licensed. A Microsoft 365 Copilot license lets you use Copilot Studio to extend Copilot with agents; otherwise your admin assigns the **Copilot Studio user license** from **Users › Active users › Manage product licenses**. Users of your agents don't need a Copilot Studio license. ## Related > [card] [Connection reference](/enterprise/connection-reference) > > Endpoint, OAuth details, and the allowlist. > [card] [Microsoft 365 Copilot](/enterprise/microsoft-365-copilot) > > Enable the built-in connector instead of building a tool. # AWS Bedrock AgentCore Gateway Source: https://docs.consensus.app/enterprise/gateways/aws-agentcore Add the Consensus MCP server as an MCP target in Amazon Bedrock AgentCore Gateway with an API key credential provider. Amazon Bedrock AgentCore Gateway treats remote MCP servers as first-class targets. Consensus's stateless server works with all supported protocol versions, including the sessionless `2026-07-28`. ## Shared API key **1. Store the key** ```bash theme={null} aws bedrock-agentcore-control create-api-key-credential-provider \ --name consensus-api-key \ --api-key "ak_live_..." ``` **2. Create the target** ```bash theme={null} aws bedrock-agentcore-control create-gateway-target \ --gateway-identifier "your-gateway-id" \ --name "Consensus" \ --target-configuration '{ "mcp": { "mcpServer": { "endpoint": "https://mcp.consensus.app/mcp" } } }' \ --credential-provider-configurations '[{ "credentialProviderType": "API_KEY", "credentialProvider": { "apiKeyCredentialProvider": { "providerArn": "arn:aws:bedrock-agentcore:::token-vault/default/apikeycredentialprovider/consensus-api-key", "credentialLocation": "HEADER", "credentialParameterName": "Authorization", "credentialPrefix": "Bearer" } } }]' ``` Or with the toolkit CLI: `agentcore add gateway-target --name Consensus --type mcp-server --endpoint https://mcp.consensus.app/mcp --gateway MyGateway`, then `agentcore deploy`. ## Per-user OAuth AgentCore's OAuth credential providers require a `clientId` and `clientSecret`; they don't perform Dynamic Client Registration. [Contact sales](https://consensus.app/home/contact/sales/) for a confidential client, then create a `CustomOAuth2` provider with `discoveryUrl` set to `https://consensus.app/.well-known/oauth-authorization-server` and use `grantType: AUTHORIZATION_CODE` on the target. The target stays in `CREATE_PENDING_AUTH` until an admin completes consent. ## Notes * The credential provider must be in the same region as the gateway. * Tool schemas are synced at create time; call `SynchronizeGatewayTargets` if the tool list changes. * Token passthrough (`JWT_PASSTHROUGH`) isn't available for MCP targets. AWS docs: [MCP server targets](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/gateway-target-MCPservers.html) · [Outbound auth](https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/gateway-outbound-auth.html) # Azure API Management Source: https://docs.consensus.app/enterprise/gateways/azure-api-management Expose the Consensus MCP server through Azure API Management as a passthrough MCP server, with a bearer token injected by policy. Azure API Management can front an external MCP server as a passthrough resource, so your agents call your APIM instance and it forwards to Consensus. Available on Developer, Basic, Basic v2, Standard, Standard v2, Premium, and Premium v2 tiers. ## Register the server **1. Create the MCP server** In the Azure portal open your API Management instance, go to **APIs › MCP servers**, select **+ Create MCP server**, then **Expose an existing MCP server**. **2. Backend MCP server** | Field | Value | | - | - | | **MCP server base URL** | `https://mcp.consensus.app/mcp` | | **Transport type** | Streamable HTTP (the default) | **3. New MCP server** Set **Name** to Consensus, a **Base path** such as `consensus`, and a **Description**. Assign **Products** and select **Create**. ## Authentication > [tab] Bearer (shared) > > Store your Consensus API key as a secret Named Value (for example `consensus-api-key`) and add an inbound policy on the MCP server: > > ```xml theme={null} > > @("Bearer " + "{{consensus-api-key}}") > > ``` > [tab] OAuth (per user) > > Use Credential manager with a confidential client from Consensus ([contact sales](https://consensus.app/home/contact/sales/); APIM doesn't do Dynamic Client Registration). Then inject the user's token: > > ```xml theme={null} > provider-id="consensus" > authorization-id="auth-01" > context-variable-name="auth-context" > identity-type="managed" > ignore-error="false" /> > > @("Bearer " + ((Authorization)context.Variables.GetValueOrDefault("auth-context"))?.AccessToken) > > ``` ## Notes * A `401` from the backend usually means the `Authorization` header wasn't forwarded; check the `set-header` policy. * Don't read `context.Response.Body` in MCP server policies; buffering the body breaks the response stream. * APIM requires the backend to speak MCP `2025-06-18` or later. Microsoft docs: [Expose an existing MCP server](https://learn.microsoft.com/en-us/azure/api-management/expose-existing-mcp-server) · [Secure MCP servers](https://learn.microsoft.com/en-us/azure/api-management/secure-mcp-servers) # Cloudflare MCP Server Portals Source: https://docs.consensus.app/enterprise/gateways/cloudflare Add the Consensus MCP server to a Cloudflare One MCP Portal so users reach it through one Access-protected URL. MCP Server Portals, part of Cloudflare One (Zero Trust), give your users a single portal URL for all approved MCP servers, with Access policies in front. The portal auto-detects Streamable HTTP and supports stateless servers. > [!NOTE] > > Cloudflare's separate **AI Gateway** product is not in the MCP path; tool calls don't traverse it. Use MCP Portals. ## Register the server **1. Open MCP Portals** In the Cloudflare dashboard go to **Zero Trust › Access controls › MCP Portals**, open the **MCP servers** tab, and select **Add an MCP server**. **2. Fill in the server** | Field | Value | | - | - | | **Name** | Consensus | | **HTTP URL** | `https://mcp.consensus.app/mcp` | **3. Choose authentication and policies** Pick one of the options below, attach your Access policies, then **Save and connect server**. ## Authentication > [tab] Bearer (shared) > > Set the authentication type to **bearer** and paste your Consensus API key as the raw bearer token (or as a custom `Authorization: Bearer ak_live_...` header). > [tab] OAuth (per user) > > Set the authentication type to **oauth** and choose **automatic OAuth registration**. Cloudflare registers with Consensus through Dynamic Client Registration and each user signs in on first use. Manual credentials also work if you prefer a confidential client: Authorization endpoint `https://consensus.app/oauth/authorize/`, Token endpoint `https://consensus.app/oauth/token/`, Scope `search`. Terraform resource: `cloudflare_zero_trust_access_mcp_server_portal`. ## Notes * Users blocked by an Access policy can still reach `mcp.consensus.app` directly unless Access is also the OAuth provider; combine with your egress policy if that matters. * Manual-OAuth servers show **Waiting** until the first user completes sign-in. Cloudflare docs: [MCP Portals](https://developers.cloudflare.com/cloudflare-one/access-controls/ai-controls/mcp-portals/) # IBM ContextForge MCP Gateway Source: https://docs.consensus.app/enterprise/gateways/ibm-contextforge Register the Consensus MCP server in IBM's open-source ContextForge gateway with a bearer token or per-user OAuth. ContextForge (`mcp-context-forge`) is IBM's open-source MCP gateway and registry. Register Consensus from the Admin UI's **Gateways/MCP Servers** tab or with the REST API. ## Shared API key ```bash theme={null} curl -s -X POST "$GATEWAY/gateways" \ -H "Authorization: Bearer $MCPGATEWAY_BEARER_TOKEN" \ -H "Content-Type: application/json" \ -d '{ "name": "consensus", "url": "https://mcp.consensus.app/mcp", "transport": "STREAMABLEHTTP", "auth_type": "bearer", "auth_token": "ak_live_..." }' ``` `MCPGATEWAY_BEARER_TOKEN` is your ContextForge admin JWT, not the Consensus key. ## Per-user OAuth ContextForge implements Dynamic Client Registration and enables PKCE automatically for authorization-code flows: ```json theme={null} { "name": "consensus", "url": "https://mcp.consensus.app/mcp", "transport": "STREAMABLEHTTP", "auth_type": "oauth", "oauth_config": { "grant_type": "authorization_code", "authorization_url": "https://consensus.app/oauth/authorize/", "token_url": "https://consensus.app/oauth/token/", "redirect_uri": "https:///oauth/callback", "scopes": ["search"] } } ``` Supply `client_id` and `client_secret` only if you've obtained a confidential client from Consensus; otherwise let the gateway register itself. ## Notes * Callers can pass their own Consensus token in `X-Upstream-Authorization`; forwarding the plain `Authorization` header requires `ENABLE_SENSITIVE_HEADER_PASSTHROUGH`. * The `transport` value is case-insensitive in recent builds; check your instance's `/docs` OpenAPI if `STREAMABLEHTTP` is rejected. IBM docs: [ContextForge on GitHub](https://github.com/IBM/mcp-context-forge) · [OAuth guide](https://ibm.github.io/mcp-context-forge/manage/oauth/) # LiteLLM MCP Gateway Source: https://docs.consensus.app/enterprise/gateways/litellm Add the Consensus MCP server to LiteLLM's config.yaml with a bearer token or per-user OAuth with Dynamic Client Registration. LiteLLM's MCP Gateway exposes upstream MCP servers on its `/mcp` endpoint and scopes them by key and team. Register Consensus in the UI under **MCP Servers › Add New MCP Server**, or in `config.yaml`. > [!WARNING] > > LiteLLM's `transport` defaults to `"sse"`. Set it to `"http"` or the connection will fail; Consensus has no SSE endpoint. ## Shared API key ```yaml theme={null} mcp_servers: consensus: url: "https://mcp.consensus.app/mcp" transport: "http" auth_type: "bearer_token" auth_value: os.environ/CONSENSUS_API_KEY description: "Search 400M+ scholarly articles, including 350M peer-reviewed papers" ``` `bearer_token` sends the value as `Authorization: Bearer `. If you'd rather set the header explicitly: ```yaml theme={null} mcp_servers: consensus: url: "https://mcp.consensus.app/mcp" transport: "http" static_headers: Authorization: "Bearer ak_live_..." ``` ## Per-user OAuth Consensus supports Dynamic Client Registration, so point LiteLLM at the registration endpoint and it registers itself: ```yaml theme={null} mcp_servers: consensus: url: "https://mcp.consensus.app/mcp" transport: "http" auth_type: oauth2 oauth2_flow: authorization_code authorization_url: "https://consensus.app/oauth/authorize/" token_url: "https://consensus.app/oauth/token/" registration_url: "https://consensus.app/oauth/register/" scopes: ["search"] ``` LiteLLM's callback is `https:///ui/mcp/oauth/callback`; set `PROXY_BASE_URL` when running behind an ingress. `oauth2_flow` is required on every `oauth2` entry. If your clients already hold a Consensus-issued token, `auth_type: true_passthrough` forwards it unchanged. ## Notes * Changing `server_id` after keys and teams have been granted access breaks those permissions. * On aggregate endpoints, per-server credentials go in `x-mcp-consensus-authorization`. LiteLLM docs: [MCP](https://docs.litellm.ai/docs/mcp) · [MCP OAuth](https://docs.litellm.ai/docs/mcp_oauth) # Obot MCP Gateway Source: https://docs.consensus.app/enterprise/gateways/obot Add the Consensus MCP server to Obot as a remote server with custom headers or spec-compliant per-user OAuth. Obot is an open-source MCP gateway and registry with per-user OAuth, a catalog, and audit logging. Every server it exposes is served as `streamable-http` at `https:///mcp-connect/`. ## Register the server **1. Add a remote server** Go to **MCP Management › MCP Servers**, select **Add MCP Server**, and choose the **Remote** type. **2. Set the URL** **Remote URL**: `https://mcp.consensus.app/mcp`. Optionally restrict connections to the hostname `mcp.consensus.app`. ## Authentication > [tab] Headers (shared or per user) > > Add a header with **Key** `Authorization` and **Value** `Bearer ak_live_...`, and mark it **Sensitive**. Set the value to **User-Supplied** instead if each user should enter their own Consensus key. > [tab] OAuth (per user) > > Nothing to configure. Consensus follows the MCP authorization spec with Dynamic Client Registration, which Obot supports out of the box; users sign in on first use. If you'd rather use a confidential client, enter the **Client ID** and **Client Secret** under static OAuth. ## Notes * Remote servers configured for static OAuth without credentials are hidden from non-admin users. * Servers using static OAuth can't be included in composite servers. Obot docs: [MCP servers](https://docs.obot.ai/functionality/mcp-servers/) · [OAuth configuration](https://docs.obot.ai/configuration/mcp-server-oauth-configuration/) # Other gateways Source: https://docs.consensus.app/enterprise/gateways/other-gateways Kong, Solo.io agentgateway, Zuplo, Docker MCP Gateway, and Boomi AI Gateway can front the Consensus MCP server with a static bearer header; each has a caveat worth knowing. These products work with Consensus but with a limitation on one of the auth modes, or with docs that are still settling. In every case the shared-API-key path is the reliable one. > [accordion] Kong AI Gateway (AI MCP Proxy) > > Use the plugin's `passthrough-listener` mode with a Gateway Service pointing at `https://mcp.consensus.app` and a route on `/mcp`: > > ```yaml theme={null} > plugins: > - name: ai-mcp-proxy > config: > mode: passthrough-listener > max_request_body_size: 1048576 > ``` > > The MCP plugin doesn't inject an upstream credential itself. Add the bearer with Kong's Request Transformer plugin: `config.add.headers` → `Authorization:Bearer ak_live_...`. Kong has no upstream OAuth broker; its OAuth plugins are inbound only. Requires Kong Gateway 3.12+ and an upstream on MCP `2025-06-18` or later. > [accordion] Solo.io agentgateway > > Add Consensus as an MCP target with a backend auth key: > > ```yaml theme={null} > mcp: > port: 3000 > targets: > - name: consensus > mcp: > host: https://mcp.consensus.app/mcp > policies: > backendAuth: > key: "$CONSENSUS_API_KEY" > backendTLS: > hostname: mcp.consensus.app > ``` > > Set `statefulMode: stateless` if you don't want the gateway to synthesize sessions. `backendAuth.passthrough` forwards a caller's own Consensus token; there is no built-in authorization-code broker. > [accordion] Zuplo MCP Gateway > > Zuplo's gateway is stateless, POST-only Streamable HTTP with no SSE, which mirrors Consensus exactly. Point the `McpProxyHandler` at the server: > > ```json theme={null} > "options": { "rewritePattern": "https://mcp.consensus.app/mcp" } > ``` > > For a shared key, add the `set-headers-inbound` policy with `Authorization` = `Bearer ak_live_...`. For per-user OAuth, the `mcp-token-exchange-inbound` policy's `clientRegistration: auto` uses Client ID Metadata Documents rather than RFC 7591 Dynamic Client Registration; use `mode: manual` with a confidential client from Consensus, or test `auto` first. > [accordion] Docker MCP Gateway and Docker Agent > > Docker's MCP Toolkit catalog doesn't document adding a remote server by URL. Docker Agent (`cagent`) does: > > ```yaml theme={null} > toolsets: > - type: mcp > remote: > url: "https://mcp.consensus.app/mcp" > transport_type: "streamable" > headers: > Authorization: "Bearer ${env.CONSENSUS_API_KEY}" > ``` > > Omit `headers` to use Dynamic Client Registration OAuth instead. Treat this as workstation tooling rather than an enterprise gateway. > [accordion] Boomi AI Gateway (formerly Lunar.dev MCPX) > > Add a remote server with `type: streamable-http` (inferred from the `/mcp` suffix) and a static header: > > ```json theme={null} > { > "consensus": { > "url": "https://mcp.consensus.app/mcp", > "headers": { > "Authorization": "Bearer ak_live_..." > } > } > } > ``` > > Boomi's docs are mid-migration from Lunar; the remote-server page isn't republished yet, and its documented static OAuth covers client-credentials and device flows only. Use the bearer path. Running one of these and hitting a snag? [Contact sales](https://consensus.app/home/contact/sales/) and we'll help verify the setup. # AI gateways and MCP registries Source: https://docs.consensus.app/enterprise/gateways/overview Register the Consensus MCP server in the gateway your organization already runs. The same five facts work everywhere; product pages give the exact field names. If your organization routes AI traffic through a gateway or MCP registry, register Consensus there once and every agent behind it gets the `search` tool. Consensus is a plain remote Streamable HTTP server, so any product that can proxy a remote MCP server can front it. ## What every gateway needs | Field | Value | | - | - | | Server URL | `https://mcp.consensus.app/mcp` | | Transport | Streamable HTTP (choose it explicitly; there is no SSE endpoint) | | Sessions | None required. The server is stateless. | | Auth, shared credential | Header `Authorization`, value `Bearer ak_live_...` | | Auth, per user | OAuth 2.0 authorization code + PKCE (S256), scope `search`, Dynamic Client Registration supported | | Discovery | `https://consensus.app/.well-known/oauth-authorization-server` | Full details, including the firewall allowlist, are in the [connection reference](/enterprise/connection-reference). ## Which auth mode to pick * **Shared API key** is the lowest-friction path and works with every product below. Everyone behind the gateway shares one Consensus account and its Enterprise limits. * **Per-user OAuth** gives each person their own Consensus account and usage. Products marked "DCR" below can register themselves automatically; products marked "client ID/secret" need a confidential client, which Consensus provisions on request. ## Product guides > [card] [TrueFoundry](/enterprise/gateways/truefoundry) > > API key · OAuth with client ID/secret > [card] [LiteLLM](/enterprise/gateways/litellm) > > Bearer · OAuth with DCR > [card] [Cloudflare MCP Server Portals](/enterprise/gateways/cloudflare) > > Bearer · OAuth with DCR > [card] [IBM ContextForge](/enterprise/gateways/ibm-contextforge) > > Bearer · OAuth with DCR > [card] [Azure API Management](/enterprise/gateways/azure-api-management) > > Bearer · OAuth with client ID/secret > [card] [AWS Bedrock AgentCore Gateway](/enterprise/gateways/aws-agentcore) > > API key · OAuth with client ID/secret > [card] [Portkey](/enterprise/gateways/portkey) > > Headers · OAuth with DCR > [card] [Obot](/enterprise/gateways/obot) > > Headers · OAuth with DCR > [card] [Kong, agentgateway, Zuplo, Docker, Boomi](/enterprise/gateways/other-gateways) > > Work with caveats Don't see your platform? Any gateway that supports remote Streamable HTTP servers with a static bearer header will work with the table above. If you'd like us to verify a setup or add a guide, [contact sales](https://consensus.app/home/contact/sales/). # Portkey MCP Gateway Source: https://docs.consensus.app/enterprise/gateways/portkey Register the Consensus MCP server in Portkey's MCP Registry with custom headers or per-user OAuth. Portkey's MCP Registry lets admins register servers once; clients then connect through `https://mcp.portkey.ai//mcp` with a Portkey API key. ## Register the server **1. Add the server** Select **MCP Registry** in the sidebar, then **Add MCP Server**. **2. Fill in the server** | Field | Value | | - | - | | **Name** | Consensus | | **Slug** | `consensus` | | **Server URL** | `https://mcp.consensus.app/mcp` | | **Server Type** | HTTP | | **Auth Type** | See below | ## Authentication > [tab] Headers (shared) > > Choose **Headers** and add `Authorization` with the value `Bearer ak_live_...`. > [tab] OAuth Auto (per user) > > Choose **OAuth Auto**. Portkey handles OAuth 2.1 with Dynamic Client Registration, so each user signs in to Consensus on first use. **Client Credentials** doesn't apply; Consensus has no client-credentials grant. ## Connect a client ```json theme={null} { "mcpServers": { "consensus": { "url": "https://mcp.portkey.ai/consensus/mcp", "headers": { "x-portkey-api-key": "" } } } } ``` Portkey docs: [MCP Registry](https://portkey.ai/docs/product/mcp-gateway/mcp-registry) · [Authentication](https://portkey.ai/docs/product/mcp-gateway/authentication) # TrueFoundry AI Gateway Source: https://docs.consensus.app/enterprise/gateways/truefoundry Register the Consensus MCP server in TrueFoundry's MCP Gateway with a shared API key or per-user OAuth. TrueFoundry's MCP Gateway registers remote MCP servers in a central registry and fronts each with a single tenant-scoped endpoint. It preserves the upstream server's transport and forwards sessions unmodified, so Consensus's stateless Streamable HTTP server needs no special handling. ## Register the server **1. Add a remote server** In the left sidebar open **MCP Gateway**, select **Add Server**, then **Connect any Remote MCP Server**. **2. Fill in the server** | Field | Value | | - | - | | **Name** | `consensus` (lowercase letters, numbers, and hyphens) | | **Display Name** | Consensus | | **Description** | Search 400M+ scholarly articles, including 350M peer-reviewed papers | | **URL** | `https://mcp.consensus.app/mcp` | | **Transport** | `streamable-http` | **3. Choose authentication** See the two options below, then save. The gateway exposes the server at `/api/llm/[tenant]/mcp/consensus/server`. ## Authentication > [tab] API key (shared) > > | Field | Value | > | - | - | > | **Authentication** | API Key | > | **Credentials** | Shared | > | **Header name** | `Authorization` | > | **Header value** | `Bearer ak_live_...` | > > Choose **Individual** instead of Shared if each user should supply their own Consensus key under the **Auth Overrides** tab. > [tab] OAuth2 Authorization Code (per user) > > Set **Authentication Type** to **OAuth2 Authorization Code**. TrueFoundry discovers the authorization and token URLs from `https://consensus.app/.well-known/oauth-authorization-server` automatically, but it requires an **OAuth2 Client ID** and **OAuth2 Client Secret**; it doesn't perform Dynamic Client Registration. [Contact sales](https://consensus.app/home/contact/sales/) for a confidential client, and register TrueFoundry's redirect URI with us: > > ```text theme={null} > https:///api/svc/v1/llm-gateway/mcp-servers/oauth2/callback > ``` > > Set **OAuth2 Scopes** to `search`. ## Notes * Headers sent in `x-tfy-mcp-headers` always override the configured auth. * Virtual Accounts can't use per-user OAuth; give them the API-key configuration. * Use the tenant-scoped URL form; the legacy `/api/llm/mcp/[server]/server` format was retired in April 2026. TrueFoundry docs: [Getting started with MCP servers](https://www.truefoundry.com/docs/ai-gateway/mcp/mcp-server-getting-started) · [Auth and security](https://www.truefoundry.com/docs/ai-gateway/mcp/mcp-gateway-auth-security) # Microsoft 365 Copilot Source: https://docs.consensus.app/enterprise/microsoft-365-copilot For Microsoft 365 admins at universities and research institutions: approve Consensus for your tenant so it appears under Sources in Copilot Chat, without opening the app store to everyone. In Microsoft 365 Copilot, Consensus is a **connector**: a source that Copilot queries live over MCP when a user asks a research question. Users find it under **Sources** in Copilot Chat and in the Researcher agent. It's not an agent, so it never appears under Agents, and in most institutional tenants users can't add it themselves. Everything under Sources is put there by an admin. This page is written for that admin. The recommended path works in tenants where the app store is closed to end users: you deploy Consensus to the people who should have it, and nothing else changes. ## Recommended: deploy Consensus for Copilot to a group [Consensus for Copilot](https://marketplace.microsoft.com/en-us/product/WA200011861) is a Microsoft 365 app, published by Microsoft, that carries the Consensus connector. Deploying it from the admin center preinstalls it for the users you choose. Microsoft's own guidance is that when user installs are blocked, admins can still deploy apps and admin-deployed apps stay available to users, so this works in locked-down tenants without changing your store policy. You need the Global Administrator role. **1. Find Consensus for Copilot in Integrated apps** In the Microsoft 365 admin center, go to **Settings › Integrated apps**. Open the **Available apps** tab and search for *consensus*. Select **Consensus for Copilot**, then **Deploy app**. ![The Integrated apps page on the Available apps tab with Consensus for Copilot listed and its availability status](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/m365-integrated-apps-get-it-now.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=b863b3accb8d731b54e3b70764aec1b9) Settings › Integrated apps › Available apps, filtered to Consensus for Copilot. If it isn't listed under Available apps, select **Get apps**, search for *Consensus for Copilot*, and select **Get it now**. Or go straight to the listing: open [Consensus for Copilot on the Microsoft Marketplace](https://marketplace.microsoft.com/en-us/product/WA200011861) and select **Get it now** there; it opens this same deployment wizard in your admin center. ![The Microsoft 365 Apps search inside Integrated apps showing the Consensus for Copilot listing by Microsoft Corporation with a Get it now button](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/m365-integrated-apps-search.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=c638bea929ddd4ec83e8df57fad3481f) Get apps opens the Marketplace inside the admin center. **2. Assign users** The **Deploy New App** wizard runs through **Configuration**, **Users**, and **Deployment**. On **Add users**, choose who gets Consensus: * **Specific users or groups** — pick the Microsoft Entra groups that should have it (a library, a department, a research institute). Recommended for a staged rollout. * **All users** — a campus-wide rollout. * **Just me** — a first test on your own account. Select **Next**. ![The Add users step showing Just me, All users, and Specific users or groups options for Consensus for Copilot](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/m365-integrated-apps-add-users.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=6812998a39a0d16d57ee197f55a61f1f) The Users step of the Deploy New App wizard. **3. Accept permissions and deploy** Accept the permission request if one is shown, then select **Next** and finish. The wizard ends on **Deployment completed** with Consensus for Copilot marked **Deployed**. Select **Done**. Deployment can take a few hours to reach every user; **View this deployment** shows its status. ![The Deployment completed screen showing Consensus for Copilot with a Deployed status and a View this deployment link](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/m365-integrated-apps-deployed.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=2a432803f6c7d2c3f77451a09b4d5d49) Deployment completed: Consensus for Copilot is deployed to the users you chose. **4. Check the two tenant gates** If a deployed user still can't see Consensus, check **Agents › Settings** in the admin center: * **Allowed agent types** must include **Allow apps and agents built by Microsoft**. The Consensus listing is published by Microsoft, so tenants that only allow their own apps will hide it. You don't need to allow external publishers. * **User access** must include the group you deployed to (**All users** or the right **Specific users/groups**). **5. Tell your users to connect** Deployment makes Consensus visible. Each user still connects once with their own Consensus account; you can't sign in on their behalf. Send them the steps in [For users](#for-users-connect-consensus), or the one-line version: *Settings › Sources › Consensus for Copilot › Connect.* > [!NOTE] > > Consensus accounts are free. Users on your institution's Consensus plan get its limits; everyone else gets the free plan. Usage never counts against your Microsoft licenses. ## Other ways Consensus appears in the admin center > [accordion] Copilot › Connectors gallery (listing being fixed) > > Microsoft lists Consensus as **Consensus (preview)** under Health and life sciences in the Connectors Gallery, but the entry doesn't currently appear when you search **Copilot › Connectors** in the admin center. Microsoft has confirmed this is a bug on their side and is fixing it. Once it shows up, you'll be able to enable it from **Copilot › Connectors › Your connections** with staged rollout to groups, the same result as deploying the app above. > [accordion] Agents › Tools > > Consensus also appears under **Agents › Tools**, Microsoft's registry of MCP servers for people who build agents. Approving it there governs which builders can use the Consensus server from Copilot Studio, VS Code, Claude Code, and GitHub Copilot CLI. Microsoft's documentation lists those as the supported surfaces; it doesn't list Copilot Chat, so don't rely on Agents › Tools to put Consensus under Sources. Use the deployment above for that. > [accordion] Custom federated connector (if your policy requires a tenant-owned connector) > > Some institutions require every connector to be created and owned in their own tenant. You can point a custom federated connector at the Consensus MCP server: **Copilot › Connectors › Gallery › Created by your org › Create a new connector › Connect to MCP server**, then: > > | Field | Enter | > | - | - | > | **Display name** | Consensus: search research papers | > | **MCP endpoint URL** | `https://mcp.consensus.app/mcp` | > | **Developer name** | Consensus | > | **Website URL** | `https://consensus.app` | > | **Privacy policy URL** | `https://consensus.app/home/privacy-policy/` | > | **Terms of use URL** | `https://consensus.app/home/terms-of-service/` | > | **Authentication type** | OAuth 2.0 | > | **Reference ID** | An OAuth client registration ID from the Teams Developer Portal | > > ![The Connect to MCP server form showing developer name, website, privacy policy, and terms fields, the Authentication type dropdown, and the Notice checkbox](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/m365-connect-mcp-form.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=f548c65692c6da1b3df8df19263cd688) > > The Connect to MCP server form. > > The Reference ID requires an OAuth client with a client secret registered in the Teams Developer Portal (**Tools › OAuth Client Registration**). Consensus's self-service registration issues public clients, so [contact us](https://consensus.app/home/contact/sales/) for a confidential client and the registration values before starting. We're also confirming with Microsoft whether a shared reference ID can be used. ## For users: connect Consensus Once your admin has added Consensus, connect it to your own Consensus account. It takes about a minute and you only do it once. **1. Open Settings** In Microsoft 365 Copilot, select the **gear icon** at the bottom of the left sidebar, then **Settings**. ![Microsoft 365 Copilot with the gear icon at the bottom of the left sidebar highlighted and the Settings menu item open](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/copilot-open-settings.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=21b881f7a4d10553a0b75488dea83392) The gear icon sits next to your name at the bottom of the sidebar. **2. Go to Sources and find Consensus** In **Chat settings**, select **Sources**. Type *consensus* in the search box to filter the list, then select **Connect** next to **Consensus for Copilot**. ![The Chat settings dialog with Sources selected and Consensus for Copilot showing a Connect button](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/copilot-sources-consensus.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=5677305cfb103c3a89f71be9696ea454) Sources lists everything your admin has made available. **3. Continue to Consensus** A dialog explains what Copilot will be able to read. Select **Continue to Consensus for Copilot**. ![The Connect Consensus for Copilot dialog with a Continue to Consensus for Copilot button](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/copilot-connect-dialog.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=ad5d94c6b7c5317d00a8a6f256cc88ce) The connection dialog before you're sent to Consensus to sign in. **4. Authorize in Consensus** A Consensus window opens. Sign in if you're not already, then select **Connect**. If you don't have a Consensus account, create a free one at [consensus.app/sign-up](https://consensus.app/sign-up) first. You'll be sent back to Microsoft automatically. ![The Consensus authorization page titled Connect Consensus to Microsoft Copilot with a Connect button](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/copilot-consent.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=46f12d15ddcae2d9c3afb543de09cf2c) The Consensus consent screen. It lists exactly what Copilot can do: search academic papers. **5. You're connected** Back in Copilot, a confirmation appears and Consensus for Copilot stays in your Sources list. Close settings and ask a research question; Copilot will search Consensus and cite the papers. Consensus is also available in the **Researcher** agent's Sources menu. ![The Sources list showing Consensus for Copilot with a toast reading Consensus for Copilot is now connected](https://mintcdn.com/consensus-f7e6ffce/iX9HFICQ-cSOT9Xi/images/enterprise/copilot-connected.png?fit=max\&auto=format\&n=iX9HFICQ-cSOT9Xi\&q=85\&s=0182b7278b13ed36d5f081024faf20dd) Consensus for Copilot is now connected. To disconnect later, open the **…** menu next to Consensus for Copilot under Sources. Results depend on your Consensus plan, not your Microsoft license. ## Why can't my users see it? | What the user sees | Likely cause | Where to fix it | | - | - | - | | Consensus isn't in the Sources list | The app isn't deployed to that user, or deployment hasn't propagated yet | **Settings › Integrated apps** › select Consensus for Copilot › **Users** | | Sources list is short and there's no search | Expected. Users only see what an admin has deployed | Deploy Consensus as above | | "Office store not available" when opening the Marketplace | Org settings block user access to the store | **Org settings › Services › User owned apps and services › Let users access the Office Store** | | The Marketplace shows Consensus but users can't install it | External or Microsoft-published apps are disallowed, or user access is set to **No users** | **Agents › Settings › Allowed agent types** and **User access** | | Connected, but searches return nothing | The user hasn't finished signing in to Consensus | Reconnect from **Sources** | | Sign-in fails with an OAuth error (custom connector only) | The Reference ID points at a registration missing PKCE, a trailing slash, or the `search` scope | Teams Developer Portal › **Tools › OAuth Client Registration** | ## Want Consensus inside your own agents? Copilot Studio can call the Consensus MCP server directly as a custom tool, with either per-user sign-in or a shared API key. See [Microsoft Copilot Studio](/enterprise/copilot-studio). Questions about tenant setup, SSO, or an enterprise plan for your organization? [Talk to sales](https://consensus.app/home/contact/sales/). # Consensus for enterprise Source: https://docs.consensus.app/enterprise/overview Consensus is used by some of the largest research organizations in the world through the MCP server and the API. Pick the deployment shape that fits your organization, then follow the guide for your platform. Consensus is used by some of the largest research organizations in the world, connected through the [MCP server](/consensus-mcp) or the [API](/api-get-started). Both give your users search over 400 million scholarly articles, including 350 million peer-reviewed research papers, from inside the tools they already use. How you wire it up depends on your organization's requirements. ## Three ways to deploy | Shape | How it works | Best for | | - | - | - | | **Per-user OAuth** | Each person connects an MCP client (Claude, ChatGPT, Cursor, and others) and signs in to their own Consensus account. No admin work. | Teams that already run AI tools and want individual usage and limits. | | **Shared gateway credential** | Your AI gateway or MCP registry holds one Consensus API key server-side and proxies requests for everyone behind it. No user redirects, no token refresh. | Organizations that route all AI traffic through a gateway (TrueFoundry, LiteLLM, Azure API Management, and others). | | **Microsoft 365 Copilot** | An admin adds Consensus as a federated connector; users connect it under **Sources** in Copilot Chat. | Microsoft 365 shops where users can't add sources themselves. | All three hit the same server, the same index, and the same [tools](/mcp-tools/quickstart). Mixing them is fine. ## Guides > [card] [Connection reference](/enterprise/connection-reference) > > Endpoint, transport, authentication options, discovery documents, and the firewall allowlist. Everything a gateway needs. > [card] [Microsoft 365 Copilot](/enterprise/microsoft-365-copilot) > > For admins: add Consensus as a federated connector so it appears in your users' Sources. > [card] [Microsoft Copilot Studio](/enterprise/copilot-studio) > > Add Consensus as a Model Context Protocol tool in your own agents. > [card] [AI gateways](/enterprise/gateways/overview) > > Register Consensus in TrueFoundry, LiteLLM, Cloudflare, IBM ContextForge, Azure API Management, AWS AgentCore, Portkey, Obot, and more. > [card] [Security and data handling](/enterprise/security) > > SOC 2 Type II compliant. What leaves your network, what comes back, and what Consensus stores. > [card] [Single sign-on](/enterprise/sso) > > Connect Microsoft Entra ID, Okta, OpenAthens, or any SAML 2.0 identity provider, with optional SCIM provisioning. ## Need help? Enterprise plans include custom rate limits, page sizes up to 1,000 papers, DOIs and full-text excerpts in results, and setup support. Consensus is SOC 2 Type II compliant and supports [SAML 2.0 single sign-on](/enterprise/sso). To set up SSO, get a confidential OAuth client for your gateway, raise limits, or get help with a platform that isn't documented here, [talk to sales](https://consensus.app/home/contact/sales/). # Security and data handling Source: https://docs.consensus.app/enterprise/security Consensus is SOC 2 Type II compliant and hosted on Google Cloud. What leaves your network when someone searches, what comes back, what Consensus stores, and the controls behind it. The questions IT and security teams ask before approving a connector. Consensus is a multi-tenant SaaS application: a web app, an MCP server, and a REST API, all hosted on Google Cloud Platform and reached over HTTPS. Nothing is installed on your premises and no VPN is required. This page summarizes the controls; the full picture is available under NDA for your vendor review. ## Compliance Consensus is **SOC 2 Type II** compliant. The audit covers the security, availability, and confidentiality of the Consensus service. Request the current report and bridge letter through [sales](https://consensus.app/home/contact/sales/); we share it under NDA as part of your vendor review. The hosting platform, Google Cloud, holds its own certifications, including ISO 27001, SOC 1/2/3, and PCI DSS. ## What is sent A search request carries the user's query and any filters (year range, study type, journal quartile, and so on), plus a credential. Requests travel over HTTPS: * **MCP server** (`https://mcp.consensus.app/mcp`): an API key or a per-user OAuth token, sent as `Authorization: Bearer`. * **REST API** (`https://api.consensus.app/v1/search`): an API key, sent in the `x-api-key` header. When your users upload documents or search their Library in the Consensus web app, that content stays in their private Library. The MCP server and API never read a user's Library; they search only the public research index. ## What comes back Publicly available academic paper metadata: title, authors, journal, year, citation count, DOI, abstract, and on paid plans a study-type classification, a key takeaway, and, for open-access papers, query-relevant excerpts from the paper's full text. Nothing in a response is derived from your organization's data. ## What Consensus stores * No personal data is stored from MCP requests. * Customer data is not used to train AI models, and the LLM providers Consensus uses retain no data from requests. * Consensus does not share user data with third parties. * Uploads to a user's Library are private to that account unless the user shares a Collection. * Enterprise agreements can include a zero-data-retention option. ## Encryption Data is encrypted in transit with TLS 1.2 or later and at rest with AES-256. Secrets and credentials are managed in Google Cloud Secret Manager. ## Infrastructure The service runs in a private Google Cloud VPC with restrictive firewall rules, across multiple zones with automated failover. Public traffic passes through Google Cloud Load Balancing and Cloud Armor, which provides WAF, OWASP protections, and DDoS mitigation. Internal traffic is segmented and TLS-encrypted. Enterprise agreements can include a contractual uptime SLA. ## Access control * Customers are logically segregated within the multi-tenant environment. * Production access is limited to authorized engineering and operations staff, requires SSO with MFA, follows least privilege, and is logged. Access grants, changes, and offboarding follow documented approval and review processes evidenced under SOC 2. * Employee devices are managed with MDM-enforced anti-malware and full-disk encryption. ## Authentication for your users * **SAML 2.0 single sign-on** with AD FS, Microsoft Entra ID, Okta, and other identity providers. Group-based access can be driven by SAML attribute mapping from your directory. See [Single sign-on](/enterprise/sso). * **Federated authentication** for academic institutions. * Institutional admins manage their own users and roles. * For MCP access, users authenticate with OAuth 2.0 (authorization code with PKCE), or your gateway holds an API key server-side and sends it as a Bearer token. See the [connection reference](/enterprise/connection-reference). * For REST API access, your service sends an API key in the `x-api-key` header. The REST API doesn't accept OAuth tokens or Bearer credentials. See [API authentication](/api-get-started#authentication). To set up SSO or SCIM provisioning for your organization, see [Single sign-on](/enterprise/sso). ## Monitoring and audit logging Access, admin, and system events are centralized in Google Cloud Logging with audit logging enabled. User sign-in and sign-out events and user actions are recorded with user attribution. Application and infrastructure monitoring runs in Datadog. Customer-specific log excerpts are available on request where contractually agreed. ## Vulnerability management CI/CD pipelines include automated dependency and misconfiguration scanning. A third-party penetration test is performed annually, and findings are triaged by severity under the vulnerability management policy. A summary is available under NDA. ## Credentials * **API keys** are shown once at creation, can be named, revoked, and rotated from the **API & MCP Dashboard**, and are scoped to search. Store them in your gateway's secret store, never in client-side code. * **OAuth tokens** are issued per user after an authorization-code flow with PKCE (S256). There is no client-credentials grant and no way to obtain a token without a user's consent. Users can disconnect at any time from their Consensus account. ## Endpoint integrity Only connect to the official endpoint, `https://mcp.consensus.app/mcp`. The OAuth discovery documents at `mcp.consensus.app/.well-known/oauth-protected-resource` and `consensus.app/.well-known/oauth-authorization-server` let a gateway verify the authorization server it is talking to. ## Contracts and data lifecycle Consensus can execute a mutual NDA and a data-processing agreement. Contracts can include terms for the return and irrecoverable disposal of your data at termination. ## Reviews and questionnaires For the SOC 2 Type II report, penetration-test summary, network topology, security questionnaires, DPAs, or SSO setup, [contact sales](https://consensus.app/home/contact/sales/). # Single sign-on (SSO) Source: https://docs.consensus.app/enterprise/sso Connect Consensus to Microsoft Entra ID, Okta, OpenAthens, or any SAML 2.0 identity provider, with optional SCIM provisioning to add and remove users from your own directory. Consensus supports single sign-on with any SAML 2.0 identity provider (IdP), including Microsoft Entra ID, Okta, and OpenAthens. Optional SCIM provisioning lets your IT team add and remove users from your own directory. ## What's supported * **Standard SAML 2.0.** Consensus is a SAML 2.0 service provider, so any IdP that speaks SAML 2.0 can connect. * **Sign-in from either side.** Users can start at [consensus.app](https://consensus.app), or select the Consensus tile in your IdP's app portal (Microsoft My Apps, the Okta dashboard). * **Direct sign-in links.** A link like `https://consensus.app/sso?domain=` sends users straight to your IdP. Useful for intranet and library pages. * **Routing by email domain.** Users with an email on your domain go to your IdP automatically. One connection can cover several domains and their subdomains. * **No pre-created accounts.** A user's Consensus account is created on their first sign-in. Existing users with the same verified email are linked to their current account, so they keep their library and history. * **Optional SCIM provisioning.** Add and remove users from your own directory. See [SCIM provisioning](#optional-scim-provisioning). ## Set up SSO Setup is a short exchange of values between our team and your IdP admin. Nothing goes live until a test sign-in works. **1. Tell us your IdP and email domains** [Contact us](https://consensus.app/home/contact/sales/) with the IdP you use and every email domain your users sign in with. We create the connection on our side. **2. Get our service provider details** We send a metadata URL, plus the Entity ID and ACS (reply) URL for manual entry. **3. Create a SAML app for Consensus** Your admin creates a SAML app for Consensus with those values, releases the [attributes listed below](#what-we-need-from-your-team), and assigns the users or groups who should have access. **4. Send us your IdP metadata URL** We finish the connection on our side. **5. Test together** We send a test link, confirm a sign-in works end to end, then turn the connection on. ## What we need from your team | Item | Details | | - | - | | **IdP metadata URL** | Holds your SSO URL, Entity ID, and signing certificate. In Entra ID this is the **App Federation Metadata URL**; in Okta it's on the app's **Sign On** tab. | | **Email domains** | Every domain your users sign in with. Users must sign in with an email on one of these domains. | | **Email attribute** | Required. We identify each user by email. | | **First and last name attributes** | Recommended. They fill in user profiles. | | **User assignment** | Only users or groups assigned to the Consensus app in your IdP can sign in. | We can map whatever attribute names your IdP already sends, so you don't need to rename them. ## Optional: SCIM provisioning SCIM lets your IT team manage who has Consensus access from your own directory. It runs on top of an SSO connection, so set up SSO first. * **Onboarding.** Users you assign in your IdP get a Consensus account before their first sign-in. * **Updates.** Profile changes in your directory sync to Consensus. * **Offboarding.** Removing or deactivating a user in your IdP deactivates their Consensus account and ends their active sessions. This happens as soon as your IdP syncs the change, usually within minutes. * **Roles.** If your team uses a Consensus organization, IdP groups can map to organization roles. To turn it on, we send a SCIM endpoint URL and a bearer token through a one-time secure link. Your admin adds them to the Consensus app's provisioning settings and turns provisioning on. > [!NOTE] > > Deactivation removes access. It doesn't delete the user's saved data. [Contact us](https://consensus.app/home/contact/sales/) if you need that on offboarding. ## Get started To set up SSO or SCIM for your organization, [contact sales](https://consensus.app/home/contact/sales/). For how Consensus handles your users' data once they're signed in, see [Security and data handling](/enterprise/security). # Go deeper with Deep Review Source: https://docs.consensus.app/get-started/deep-review A Deep Review runs many searches, reads 50 to 100 papers, and writes a longer report. Use it when one search isn't enough to answer your question. A normal search (**Pro**) reads up to 20 papers and gives you a focused answer in seconds. A **Deep Review** does what a researcher would do over an afternoon: it runs up to 20 searches at once, follows citations, and reads 50 to 100 papers before writing a longer report. ## When to use it | Use Pro when… | Use Deep when… | | - | - | | You have a specific question | Your question is broad or has several parts | | You want a quick answer | You want an overview of a whole topic | | You're exploring | You're checking an idea or looking for gaps | A good rule: start with Pro. If the answer feels thin, or you find yourself asking five follow-ups, switch to Deep. ## Run one **1. Turn on Deep** Click **Deep** in the search bar before you send your question. It turns blue when it's on. You can also click it on an answer you already have to upgrade it. ![The Consensus search bar with the Deep toggle switched on and highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/deep-toggle-on.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=66908db5a6f87b4ac22c98f390e9bd2a) **2. Ask a big question** Deep does best with questions that need many papers to answer well. ```text theme={null} What does the research say about treatments for endometriosis pain? Compare hormonal, surgical, and non-drug approaches, and note where the evidence is weak or conflicting. ``` **3. Wait a few minutes** Deep takes a few minutes. You can watch its plan in the side panel, or leave and come back. Consensus emails you when it's done. **4. Read the report** The report is longer than a Pro answer and includes visuals such as a summary of where the evidence leans, key claims and their evidence, and research gaps. Every sentence is cited. > [!NOTE] > > Deep Reviews are limited by plan: 3 per month on Free, 15 on Pro, and 200 on Deep. Save them for your big questions. ## Try it on your topic Run one Deep Review on the broadest version of your topic. Skim the report for the main themes and any gaps it mentions. > [!TIP] > > **You should now have:** a Deep Review report on your topic that you can come back to as you work through the rest of this guide. > [accordion] Go deeper: how Deep Review works > > See [Deep Review](/core-features/deep-search) for what happens at each step, example prompts, exporting the report, and running a Deep Review over your own papers. # Read the full paper Source: https://docs.consensus.app/get-started/full-text-access Consensus reads the whole paper, not just the abstract, including paywalled articles from major publishers, open-access PDFs, and papers you upload yourself. [Full-text search for paywalled papers - built into Consensus](https://www.youtube.com/embed/-k161sWri8Q) An abstract is the short summary at the top of a paper. It tells you what the study found, but not how well it was done: the sample size, the methods, and the limitations are in the body of the paper. Most AI tools only read abstracts. Consensus reads the full text whenever it can, including many paywalled papers, so its answers can draw on the methods, results, and discussion. You can also click through to the exact passage it used. ## Where full text comes from > [accordion] Publisher partnerships > > Consensus has agreements with the publishers below to read the full text of paywalled papers for AI analysis inside Consensus, including 7 of the 10 largest academic publishers. We're constantly adding more publishers to this list. > > ![Wiley](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/wiley.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=89f93113e74bb0090ce94a5b4f55c24a) > > ![Taylor & Francis](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/taylor-and-francis.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=2075ac1161cb69642ac53438254b2aae) > > ![SAGE](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/sage.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=da227aff27e194e0395d10dddb992ece) > > ![Wolters Kluwer](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/wolters-kluwer.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=8d67fdc3b4fa013aa8cc8429b49de5f7) > > ![Oxford University Press](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/oxford-university-press.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=4159326ee81748817a0c41594c71c17d) > > ![American Chemical Society](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/acs.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=6b8e12edf4b83daef036c2aa4b8c14a9) > > ![De Gruyter Brill](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/de-gruyter-brill.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=efed8f09bb4ea863bfa97cc33566c933) > > ![BMJ](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/bmj.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=73f5754cbec30c1ad3ef2ce20a71d81c) > > ![American Psychological Association](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/apa.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=29cfd1c4ea69e51519e5a8edda1081a5) > > ![AAAS](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/aaas.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=6ebd3cc659c6f52f1f4e95d8af2b8aec) > > ![JAMA Network](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/jama.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=d9a7d8740a81df20c319564a003723bd) > > ![American Society for Microbiology](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/asm.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=b0ea127517100a30f2a38ca56eb60467) > > ![Rockefeller University Press](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/logo/publishers/rockefeller-university-press.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=a2a1689ee6762e8ae794f5578e7734e4) > >
    >
  • Wiley
  • >
  • Taylor & Francis
  • >
  • SAGE Publishing
  • >
  • Wolters Kluwer
  • >
  • Oxford University Press
  • >
  • American Chemical Society (ACS)
  • >
  • De Gruyter Brill
  • >
  • BMJ
  • >
  • American Psychological Association (APA)
  • >
  • American Association for the Advancement of Science (AAAS)
  • >
  • JAMA Network
  • >
  • American Society for Microbiology (ASM)
  • >
  • Rockefeller University Press
  • >
> [accordion] Open-access PDFs > > For any open-access paper, Consensus pulls the full PDF directly into the app. When the AI quotes from it, you can click straight to the passage in the PDF. > [accordion] Your own uploads > > If a paper isn't in the index, bring it. Upload PDFs directly or sync your Zotero library. Uploads stay private to your account and are analyzed with the same engine, including full-text search, snippet highlighting, and synthesis. See [Library](/core-features/library). > [!NOTE] > > **What about my institution's subscriptions?** Consensus connects to your university or institution's library through LibKey, but only for link resolving: it gives you direct links to papers your library subscribes to. Consensus doesn't use full text from LibKey. All the full text Consensus analyzes comes from the three sources above. Full-text availability varies by paper. When it isn't available, Consensus relies on the abstract and indexed metadata. ## Where full text is used * **Retrieval.** Your query is matched against titles, abstracts, and full text, which is one reason Consensus retrieval tests 10% more accurate than Google Scholar head to head. * **Pro and Deep analysis.** Answers pull directly from full text when available. A document icon with a tick next to an in-line citation means the full text was used; no icon means the abstract was. * **Citation Grounding.** Hover a citation to see the exact quote; click to open the paper at that passage. See [Citation Grounding](/core-features/citation-grounding). * **Comparison tables.** Side-by-side values for methods, sample sizes, findings, and limitations, pulled from the body of each paper. * **Chat with full text.** Ask questions of one paper, a Collection, or your whole Library. > [!NOTE] > > Analysis access and reading access are different. Consensus may use a paywalled paper's full text to generate summaries and tables, but the article itself may remain behind the publisher's paywall. Viewing or downloading it depends on your institutional or personal subscription. > [!NOTE] > > **Full text in the API and MCP.** The [Consensus API](/api-get-started) returns query-relevant full-text excerpts from open-access papers alongside each result, in the `full_text_chunks` field (see the [`/v1/search` reference](/api-reference/search)). The [MCP server](/consensus-mcp)'s `search` tool returns them too, so your assistant can quote the relevant passages directly. Both are available on all paid plans (Pro, Deep, Teams, and Enterprise). ## How to use it **1. Look for the full-text icon** In any Pro or Deep answer, citations with a small document-and-tick icon were analyzed from full text. Hover one and its card says **Full Text used**. **2. Jump to the passage** Hover a citation to see the supporting quote. To read it in context, click the paper's **Supporting Quotes**, open the **Attachment** tab to see the PDF, and pick a quote from the menu at the bottom to highlight it. ![The paper panel on its Attachment tab showing the PDF, with the Select a quote to highlight menu open](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/paper-attachment-quotes.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=5fd230c05fd53be52586ae85acdd0be2) Pick a quote and Consensus jumps to it and highlights it. **3. Chat with a paper** Click **+** in the search bar and select papers from your Library or upload new ones, tick papers in a results list, or ask from a paper's details page. Chatting with full text runs in Pro or Deep mode. **4. Filter to open access** Use the **Open access** filter to see only papers whose full text is viewable. **5. Connect your institution** In **Settings › Preferences**, search for your institution to get LibKey links to papers your library subscribes to. This only changes where links take you, not what Consensus analyzes. If a paper is still paywalled, try your library portal directly. ## Availability Full text is used wherever it's available, on every plan. Chatting with the full text of a paper, Collection, or Library requires a Pro or Deep message, which draws on your plan's credits (see [What is Consensus](/get-started/what-is-consensus#plans)). On all paid plans, the API and MCP also return excerpts from open-access papers alongside each result. ## Try it on your topic Run a search on your topic, then find one citation with the full-text icon and one without. Hover each to see where the quote came from. > [!TIP] > > **You should now know:** how to tell whether an answer used a paper's full text or only its abstract, and how to jump to the passage in the paper. # Paper search Source: https://docs.consensus.app/get-started/how-search-works Consensus finds the most relevant papers with hybrid keyword and semantic search, ranks them the way a researcher would weigh them, refines the top 20 with a larger model, and only then uses AI to synthesize an answer. Consensus puts discovery at the center of everything it does, combining classic and modern search techniques. It only uses AI after it searches the scientific literature, and that ordering is what makes every answer traceable to a real paper. Here's the pipeline. ## Step 1: Cast a wide net Consensus scans its entire index of 400M+ scholarly articles and matches your query against titles, abstracts, and full text (when available) using **hybrid search**: classic keyword search combined with AI-powered semantic search. * **Semantic search**, powered by AI embeddings, captures the intent behind your question, so conversational prompts and natural-language questions work. * **Keyword search**, powered by BM25, anchors results to the exact terms in your query. * **Boolean operators and exact matching** work too, for when you need precise control. ![Diagram showing Semantic Scholar, OpenAlex, and Consensus's own crawl feeding into the Consensus index](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/data-sources.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=2cc015490ef0abe6c5cdbf72af24b099) The index combines Semantic Scholar, OpenAlex, Consensus's own crawl of the scholarly web, and publisher partnerships. Together they assign a relevance score to every paper. Consensus also rewrites typos, expands MeSH synonyms, and supports Boolean operators, so keywords, questions, and structured queries all work. ## Step 2: Refine by quality The top 1,500 papers are re-ranked. The ranking logic is designed to mimic how a researcher weighs research: * **About 80% is textual relevance**: how well the paper actually matches your question. * **About 20% is research-quality signals**, such as: * **Citation count**: how often other researchers have cited the paper. Influence, not proof of validity. * **Recency**: how recently it was published, so new work can surface alongside established papers. * **Journal reputation**: the quality and impact of the journal it was published in. Relevance comes first, so a well-cited paper that only loosely matches your question won't crowd out the papers that answer it. ## Step 3: Rank the top 20 for precision Consensus recalculates textual relevance on the top 20 papers using a larger, more compute-intensive model optimized for precision rather than broad recall. Quality signals still factor in, so the final list reflects both relevance and rigor. ## Step 4: Check relevance before writing anything Before any AI summary is generated, separate checker models verify that each paper actually contains information relevant to the query. Papers that don't meet the threshold are excluded. If nothing relevant is found, Consensus tells you rather than letting a model fill in gaps that don't exist. ## Step 5: Analyze and synthesize Once relevant papers are retrieved, AI is used in two ways: * **Analyze individual papers** — for example, the Study Snapshot, which extracts population, sample size, duration, location, methods, outcomes, and results. * **Synthesize across papers** — the Pro answer, the Consensus Meter, claims and evidence tables, research gaps heat maps, and timelines. Models are instructed and trained to generate responses only from the text of the retrieved papers. Consensus uses commercial models (like OpenAI's) for general summarization and fine-tuned open-source models for domain-specific features like the Consensus Meter. In Pro and Deep modes, [Research Agent](/core-features/research-agent) adds planning on top of this pipeline: it breaks a complex question into sub-searches, applies filters, walks the citation graph, reads across papers, and composes a citation-backed answer. [Deep Review](/core-features/deep-search) is the extended version: up to 20 searches in parallel, multiple agent tools like citation crawl, and 50 to 100 papers read before answering. ## How we measure search quality Consensus tests its search head to head against the tools researchers already use: * **Google Scholar:** Consensus beat Google Scholar on retrieval accuracy by 10%. * **PubMed:** Consensus beat PubMed on medical retrieval accuracy by 9%, as judged by third-party clinicians. ## Why hallucinations are contained Consensus only searches its index of academic literature, and every source it cites is tied to a real academic paper. On top of that, Consensus has a number of guardrails built in to prevent hallucinations. The most important is [Citation Grounding](/core-features/citation-grounding), which shows you the exact sentence behind each claim and exactly where in the paper it comes from. ![A citation card over a Consensus answer, with the Full Text used badge and the supporting quote highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/citation-card.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=f667ec73080c1427293b47e81dd8c534) Hover a citation to see the exact quote behind it. > [!NOTE] > > Consensus is built to help researchers find relevant papers faster. As always, it's important to read the primary source material. Consensus has many features that show you exactly what's happening at each step of the process and exactly where in a paper information comes from, so you can save time and go straight to the paper itself. ## Next > [card] [Read the full paper](/get-started/full-text-access) > > How full text sharpens retrieval and analysis. > [card] [Start here](/get-started/what-is-consensus) > > The guide map and video playlist. # Run your first search Source: https://docs.consensus.app/get-started/run-your-first-search Type a research question, pick what to search, and read what comes back. Keyword searches and full questions both work. The fastest way to learn Consensus is to search something you care about. This lesson uses endometriosis as the example topic. Swap in your own. ## Step 1: Type your search Go to [consensus.app](https://consensus.app) and type in the search bar. You can search two ways: | Search style | Example | When it helps | | - | - | - | | Keywords | `endometriosis pain treatment` | You know the topic but not the question yet | | A question | `What treatments reduce pain in endometriosis?` | You want a direct answer, with the evidence behind it | Both work. Consensus matches the meaning of your search as well as the exact words, fixes typos, and understands medical synonyms. A clear question usually gets a more focused answer, so start with keywords if you're exploring and switch to a question once you know what you're asking. > [!TIP] > > Write the way you'd ask an expert colleague. "Does exercise help with depression in teenagers?" is better than "exercise depression". ## Step 2: Choose what to search (optional) The search bar has a few controls. You can leave them all alone for your first search. ![The Consensus search bar with the +, Skills, Deep, and Filter controls highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/search-bar.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=1318640978af8e656f1ad114f6b7969e) * **+** adds sources. Attach papers or Collections from your Library, upload a PDF, connect Zotero, or open **Search corpus** to choose which papers are searched: **All** (400M+ papers), **Medical** (top medical journals and clinical guidelines), or **My Library** (papers you've saved). * **Skills** runs a ready-made research workflow, such as a structured literature review. See [Skills](/core-features/skills). * **Deep** switches on a much longer review. With it off, your search runs as a normal **Pro** search that reads up to 20 papers. See [Go deeper with Deep Review](/get-started/deep-review). * **Filter** narrows results by year, study type, journal quality, and more. You can also just write filters into your search: "...published since 2020, RCTs only". ## Step 3: Read what comes back ![A Consensus search about endometriosis pain treatments, with the written answer and a Yes/Possibly/Mixed/No meter on the left and a list of reference papers on the right](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/results-page.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=fa7a9c1962d241a9e2ebebd57e9a4b32) The answer is on the left. The papers it used are on the right. * **At the top of the answer** you'll see the steps Consensus took, like **Pro · 3 steps**: what it searched for, how many papers matched, and how many it read. Click it to see the details. * **The answer** summarizes what the research says. For a yes/no question you'll also see a meter showing how many papers lean yes, possibly, mixed, or no. The grey tags in the text (like `Sun 2026`) are citations. * **The References panel** lists the papers behind the answer, each with a one-line key takeaway, its supporting quotes, a study-type badge, the year, and how often it's been cited. The next group, [Understand your results](/learn/read-a-consensus-answer), walks through all of this in detail. ## Step 4: Ask a follow-up Type in the **Ask a follow up** box under the answer. Consensus remembers the conversation, so you can narrow it down without starting over: ```text theme={null} Only include studies from the last 5 years. ``` ```text theme={null} Which of these are randomized controlled trials? ``` ## Try it on your topic **1. Search with keywords** Type 2–4 keywords about your topic and look at what comes back. **2. Rewrite it as a question** Turn it into a question you actually want answered and search again. Compare the two answers. **3. Ask one follow-up** Narrow it by year, study type, or population. > [!TIP] > > **You should now have:** one search on your own topic, phrased as a question, with at least one follow-up. > [accordion] Go deeper: advanced search tricks > > * Search for a specific paper with its DOI or title, or for a researcher by name. > * Boolean operators (`AND`, `OR`, `NOT`) work if you're used to them. > * See every filter in [Filter for strong, recent studies](/search-basics/find-high-quality-recent-papers#every-filter-explained). > * See what happens behind the scenes in [Paper search](/get-started/how-search-works). # Save papers to your Library Source: https://docs.consensus.app/get-started/save-papers-to-your-library Bookmark papers and searches as you go, and keep them in a Collection for your project so nothing gets lost. You'll find dozens of useful papers in your first week. **My Library** keeps them in one place, so you aren't relying on open tabs or your browser history. It's free on every plan and private to you. ## Step 1: Save a paper With a search open, hover any paper in the **References** panel and click the **bookmark** icon. That saves the paper to **My Library** straight away, and opens a menu where you can also add it to a Collection or create a new one. It's easy to miss, but it's the quickest way to keep a paper you want to come back to. ![The References panel with a paper hovered, showing its action icons with the Save paper bookmark highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/save-from-references.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=d02671b9d77385847c8501f6d61f0006) Hover a paper in References and click the bookmark to save it. The same menu appears when you click **Save** at the bottom of a paper's details panel (open it by clicking the paper's **Supporting Quotes**). Tick a Collection to file the paper there, or type in the search box to find one. ![The Save menu with My Library ticked and a search box for adding the paper to a Collection](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/save-menu.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=e76aae5a5353d5737f0e35be0fdce8aa) To save the whole search, including the answer, click the bookmark at the top of the thread. ## Step 2: Make a Collection for your project A **Collection** is a folder of papers. Make one for your research project and save everything relevant to it there. ![My Library with the Collections list in the left sidebar and the New Collection button highlighted](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/my-library-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=46213478fef550ff76bc60b62ce45f65) Collections live in the left sidebar of My Library. ## Step 3: Find it again Open **My Library** from the left sidebar in Consensus. Click a Collection to see its papers, and sort them by title, author, year, or study design. ## Try it on your topic Create a Collection named after your project and save at least 5 papers from your searches so far. > [!TIP] > > **You should now have:** a Collection for your project with your first papers in it. You'll build on it in [Organise and find gaps](/learn/build-collections). > [accordion] Go deeper: more Library features > > Upload PDFs, import from Zotero, share a Collection, export a bibliography, and ask questions of your saved papers. See [Library](/core-features/library). # Start here Source: https://docs.consensus.app/get-started/what-is-consensus A short, video-first guide to Consensus for researchers at any level. Learn to search, read, and organise the literature on your topic, one lesson at a time. [Consensus 101](https://www.youtube.com/embed/aqLYSofgv80) Consensus is an AI search engine for research. You ask a question, it searches over 400 million scholarly articles, including 350 million peer-reviewed research papers, and it writes an answer where every sentence links back to the paper it came from. It's a search engine first. It finds the papers, then uses AI to read and summarize them. It doesn't make things up from general knowledge, so you can always click through and check the source. ## How this guide works This guide is for researchers at any level, whether you're starting your first project or you've been publishing for years. Each lesson is a short video plus a prompt to try on your own topic. Start at the top if you're new to Consensus, or jump straight to the task you need. > [card] [1. Quick start](/get-started/run-your-first-search) > > Run a search, see where the papers come from, and save what you find. > [card] [2. Understand your results](/learn/read-a-consensus-answer) > > Read an answer, check its claims, and find the strongest papers. > [card] [3. Map the literature](/search-basics/map-field-history) > > Find the main findings, the history, and where studies disagree. > [card] [4. Follow the trail](/search-basics/trace-citations) > > Use citations to go from one good paper to many. > [card] [5. Organise and find gaps](/learn/build-collections) > > Keep your papers in one place and spot what's missing. > [card] [Key features](/get-started/how-search-works) > > Reference pages for search, Research Agent, Deep Review, Library, and more. All the videos are also in one [YouTube playlist](https://www.youtube.com/playlist?list=PL5at_-ZVRXUFdJpolkZcVHKmQBIGB8RXL). > [!TIP] > > New to research terms like "meta-analysis" or "preprint"? Keep [Research terms in plain English](/learn/research-terms) open in another tab. ## Where you can use Consensus > [card] [Web app](https://consensus.app) > > Everything in this guide happens at consensus.app. It's free to start. > [card] [In ChatGPT, Claude, Copilot, or the API](/core-features/mcp) > > Connect Consensus to the AI assistant you already use in one click, or build on it with the API. ## Why researchers trust Consensus * **Search happens before AI.** The AI only summarizes papers Consensus actually found. If nothing relevant turns up, it tells you. * **Every claim is checkable.** Hover a citation to see the exact quote from the paper. See [Check a claim against the paper](/core-features/citation-grounding). * **Retracted papers are flagged** and never used in answers. * **Your data isn't training data.** Consensus doesn't train models on your searches or uploads. * **It's measured.** In blinded evaluations judged by PhD researchers, Consensus finds more relevant papers than Google Scholar and than frontier models with web search. See [Benchmarks](/benchmarks). > [card] [More relevant than Google Scholar](/benchmarks#consensus-vs-google-scholar) > > 88.1% vs 81.8% precision on 500 real user queries, judged blind by PhD researchers. > [card] [Beats GPT-5.5 and Opus 4.8, 10x faster](/benchmarks#consensus-vs-frontier-models) > > More relevant top-20 papers at a median of 8.7 seconds, versus 75 to 103 seconds. > [!NOTE] > > AI is a research assistant, not a replacement for reading. Use it to find and understand papers faster, then read the ones that matter. ## Plans | Plan | Pro messages | Deep Reviews | API & MCP calls | | - | - | - | - | | Free | 10 per month | 3 per month | 30 per month | | Pro (\$20/mo, \$144/yr) | Unlimited | 15 per month | 500 per month | | Deep (\$65/mo, \$540/yr) | Unlimited | 200 per month | 2,000 per month | | Teams | Unlimited | 50 per user per month | 500 per user per month | Paper search is unlimited on every plan. See [pricing](https://consensus.app/pricing) for details, and [help.consensus.app](https://help.consensus.app) for account help. # Where the papers come from Source: https://docs.consensus.app/get-started/where-papers-come-from Consensus has the second-largest research index in the world: abstracts and metadata for 400 million scholarly articles, including 350 million peer-reviewed research papers, open-access full text, and paywalled full text through publisher partnerships, updated weekly. Consensus has the second-largest research index in the world, after Google Scholar. It searches over 400 million scholarly articles, including 350 million peer-reviewed research papers. The index covers every domain of science, from astronomy to sociology to biochemistry, and is updated weekly. ## Sources The index is built by aggregating a wide range of sources in three layers: ![Pyramid showing three layers: abstracts and metadata across all publishers from Semantic Scholar, OpenAlex, and PubMed; all open-access full text; and full text of paywalled papers from partner publishers](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/papers-pyramid.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=c6dd63f78277c9e6af4af8144e9c65a9) Abstracts and metadata for every paper, full text for open-access papers, and full text for paywalled papers through publisher partnerships. > [accordion] 1. Abstracts and metadata across all publishers > > Every paper in the index has its abstract and metadata, regardless of publisher. This layer comes from: > > * **Semantic Scholar**, which draws on over 50 direct partnerships with publishers, data providers, and aggregators covering more than 500 journals, university presses, and scholarly societies. > * **OpenAlex**, an open database that connects papers, authors, institutions, journals, and topics worldwide. > * **PubMed**, over 37 million biomedical and life-science records. > * **Consensus's own crawl of the scholarly web**, which fills the coverage gaps the aggregators miss. > [accordion] 2. All open-access full text > > Consensus pulls in the full text of every open-access paper. That lets it read the whole paper, not just the abstract, when searching and synthesizing, and lets you see exactly where in the paper each piece of information comes from. > [accordion] 3. Access to the full text of paywalled papers > > Partnerships with publishers allow Consensus to use the full text of paywalled papers in its search and synthesis. We provide a link to the paper page. Consensus doesn't replace your subscription; it complements it. > > Consensus partners with 7 of the 10 largest academic publishers. Partners include Wiley, Taylor & Francis, SAGE, Wolters Kluwer, Oxford University Press, the American Chemical Society, De Gruyter Brill, BMJ, the American Psychological Association, AAAS, JAMA Network, the American Society for Microbiology, and Rockefeller University Press, with more on the way. See [Full text access](/get-started/full-text-access). Together, these layers cover nearly all of the highest-impact journals and the entirety of PubMed. ## What's in the index * **400M+ scholarly articles in total.** About **350 million** are peer-reviewed, published research papers. The other **50 million** or so are other scholarly sources, such as books and conference proceedings. * **Preprints are included and labeled.** Use the **Exclude preprints** filter (or `exclude_preprints` in the API and MCP) to keep only peer-reviewed work. * **Retracted papers are marked** with a RETRACTED badge and are never used in AI analyses or summaries. * **Q1 journal coverage** across disciplines, including Nature, Cell, NEJM, JAMA, American Economic Review, Econometrica, and Chemical Reviews. Q1 means the top 25% of journals in a field by SCImago Journal Rank. ## You can control what you search ### Select your corpus Click **+** in the search bar, then **Search corpus**, to choose what Consensus searches. ![The + menu open on Search corpus, showing the All, Medical, and My Library options](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/search-corpus-menu.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=0705f2a47bbaee224bef88de23eef4f5) | Corpus | What it searches | | - | - | | All | The complete index of 400M+ scholarly articles | | Medical | A curated subset of \~8 million papers and 50,000 clinical guidelines from the top 1,000 medical journals, selected using Impact Factor, SJR score, and SciScore's Rigor and Transparency Index. Examples: NEJM, Cochrane, JAMA, BMJ, Circulation, Nature Medicine. | | My Library | Papers you've imported from Zotero or uploaded yourself. Search a specific Collection or your whole [Library](/core-features/library), on its own or combined with the broader index in the same search. | ### More control over the papers that surface Consensus has a number of proprietary filters built for researchers. These include: * **19 study types**, from RCTs and cohort studies to systematic reviews and meta-analyses * **Country**, where the study population is from * **Journals**, by SCImago journal rank * **Publishers**, with 44 to choose from Plus publication year, citation count, sample size, study duration, human or animal setting, open access, and preprints. Queries can also include DOIs, author names, paper titles, MeSH synonyms, and Boolean operators. > [card] [Every filter, explained](/search-basics/find-high-quality-recent-papers#every-filter-explained) > > The full list of filters, what each one does, and how to apply them. ## What happens when you hit search > [accordion] The short version > > 1. **Cast a wide net.** Hybrid search matches your query against all 400M+ articles, by meaning (AI semantic search) and by exact keywords. Natural-language questions, Boolean operators, and exact matches all work. > 2. **Rank by quality.** The top 1,500 are re-ranked the way a researcher would weigh them: about 80% on how well each paper matches your question, and about 20% on quality signals like citation count, recency, and journal reputation. > 3. **Pick the best 20.** A larger model re-reads the top papers for precision. > 4. **Check relevance.** Separate models confirm each paper actually addresses your question. If none do, Consensus says so. > 5. **Write the answer.** Only now does AI summarize, using only those papers. The full walkthrough, step by step, is in [Paper search](/get-started/how-search-works). > [!TIP] > > **You should now know:** what Consensus searches, and how to switch between **All**, **Medical**, and **My Library** under **+** › **Search corpus**. # Ask a paper Source: https://docs.consensus.app/learn/ask-questions-about-a-paper Attach a paper to the search bar and ask about its methods, sample, findings, and limitations, so you understand a study before you cite it. Once you've found an important paper, you need to understand it: who was studied, what was done, what was found, and what the weaknesses are. Consensus can read the paper with you and answer questions about it, pointing to the exact passage for each answer. ## Step 1: Attach the paper There are three ways: * **From a citation.** Hover a citation in the answer and click **Ask**. * **From your results.** Tick the box next to a paper in the References panel. * **From the search bar.** Click **+**, then **Attach papers or collections** to pick papers from your Library, or **Upload papers** to add a PDF. ![A citation card over the answer, showing the paper title, a supporting quote, and the Ask button highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/citation-card-ask.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=753eba4ab42d9fd64cb7fb03a7e8ac12) Hover a citation and click Ask to attach that paper. The paper appears above the search bar and is ticked in the References panel. You can attach several papers at once to compare them. ![A Consensus answer with one paper attached above the search bar, the same paper ticked in the References panel, and a bar offering Save and Extract to table](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/attach-paper.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ca816a6284ecf8a149689516ff514bcf) The attached paper sits above the search bar. Ticked papers can also be saved or extracted to a table. ## Step 2: Ask about it Questions that work for almost any study: ```text theme={null} What question was this study trying to answer, and why? ``` ```text theme={null} Who was studied, how many people, and for how long? ``` ```text theme={null} What was the main finding, and how big was the effect? ``` ```text theme={null} What are the main limitations of this study, according to the authors? ``` ```text theme={null} What do the authors say future research should look at? ``` ## Step 3: Check the answer against the paper Hover a citation in the answer to see the quote it came from. To read it in context, click the paper's **Supporting Quotes**, open the **Attachment** tab to see the PDF, and pick a quote from the menu at the bottom to highlight it. Read the surrounding paragraph yourself. This is the habit that makes AI a learning tool rather than a shortcut. > [!TIP] > > Clicking a paper's **Supporting Quotes** opens a paper panel. Its **Snapshot** tab pulls out the population, sample size, methods, and results at a glance, and its **Evidence** tab lists every quote the answer used. ## Try it on your topic Attach one of the key papers from [Read a Consensus answer](/learn/read-a-consensus-answer) and ask the five questions above. > [!TIP] > > **You should now have:** a 3-line summary of one key paper in your own words: what they did, what they found, and the main limitation. > [accordion] Go deeper: full text and paywalls > > Answers are best when Consensus has the full text. See [Read the full paper](/get-started/full-text-access) for where full text comes from, and how to connect your university library for links to papers it subscribes to. # Build Collections for your project Source: https://docs.consensus.app/learn/build-collections Organise your project's papers into Collections by theme, add papers from anywhere, ask questions of everything you've saved, and share it with collaborators. In [Save papers to your Library](/get-started/save-papers-to-your-library) you made a Collection for your project. By now it probably holds a few dozen papers. This lesson turns it into a reading list you can work from and share. ## Step 1: Split it by theme Collections can be nested. Create a sub-Collection for each theme you found in [Spot the main findings and themes](/learn/spot-main-findings): click the new-Collection button and pick your project Collection under **Create in**. ![My Library with the Collections list in the left sidebar and the New Collection button highlighted](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/my-library-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=46213478fef550ff76bc60b62ce45f65) Collections and sub-Collections live in the left sidebar of My Library. ## Step 2: Add papers from anywhere Click **+ Add** in the top right of My Library to: ![My Library with the + Add menu open, showing Upload, Paste DOIs, and Import from Zotero](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/library-add-menu.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=5318d155b8f455ddcd2c5cf49558ca8c) * **Upload PDFs**, such as papers a colleague sent you * **Paste DOIs** from a reading list * **Import from Zotero**, if you already use it ## Step 3: Ask questions of your Collection Open the Collection and type in the search bar at the bottom. Consensus answers using only the papers in that Collection. ```text theme={null} Which papers in this collection support the claim that [statement]? ``` ```text theme={null} Summarize what this collection says about [topic], and which papers I should read first. ``` ## Step 4: Share it with collaborators **1. Open the Collection** In **My Library**, click the Collection in the left sidebar. The **Share** button appears in the top right. ![A Collection open in My Library, with the Share button in the top right highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/share-collection-01.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ab61c1adbfdbc0383aa1cc63d73ce496) **2. Share by email or link** Click **Share**. Type an email address under **Invite via email** to invite someone directly, or click **Copy invite link** to share a link that lets anyone who has it join as a viewer. **Viewers** can read the papers and search the Collection; **Editors** can also add and remove papers. People without a Consensus account are prompted to create a free one. ## Try it on your topic Split your project Collection into sub-Collections by theme, then share it with a collaborator. > [!TIP] > > **You should now have:** an organised, shareable Collection that someone else can open and understand. > [accordion] Go deeper: Zotero, export, and more > > See [Library](/core-features/library) for Zotero import details, file limits, exporting a bibliography or RIS file, and running a Deep Review over a Collection. # Explore the citation graph Source: https://docs.consensus.app/learn/explore-the-citation-graph See how papers connect through citations on a visual map. Start from a paper you trust and find the work it built on and the work that followed. [First Look: Citation Graph Beta](https://www.youtube.com/embed/veNfFAEbTJs) Following citations one paper at a time works, but it's slow. The **Citation Graph** shows the connections all at once: start from a paper you trust (a **seed**) and see the papers around it. ## Step 1: Open a graph from a paper Open one of your key papers and click **Citation Graph** in its sidebar. You can also click **Graph** in the left sidebar and search for a paper to use as a seed. ## Step 2: Read the map ![A citation graph with seed papers and their clusters connected by citation arrows](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/citation-graph-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=fa81f851d9a177afa1055c601ebfc3e4) Older papers on the left, newer on the right. Bigger circles are more cited. * **Left to right is time.** Papers on the left came before your seed; papers on the right came after. * **Bigger circles** have more citations. Big circles on the left are often the field's foundations. * **Arrows** point from a paper to the paper it cites. Hover a circle to see its connections. Click it to see the paper's details. ## Step 3: Ask the graph a question Click **Chat** at the bottom of the page. Every paper on the graph is attached, so you can ask: ```text theme={null} How did the later papers build on the earliest one? ``` ```text theme={null} Which of these papers used a randomized design? ``` ## Try it on your topic Build a graph from your most important paper. Find one older paper it built on and one newer paper that built on it. > [!TIP] > > **You should now have:** 2–3 new papers for your reading list that you didn't find through search. > [accordion] Go deeper: seeds, density, and ranking > > Add up to 5 seeds, change how many papers are shown, promote a paper to a seed, and learn how papers are ranked. See [Citation Graph](/core-features/citation-graph). # Find research gaps Source: https://docs.consensus.app/learn/find-research-gaps Ask Consensus what's missing from the papers you've saved, such as populations, study types, or topics, then search the full index for papers that fill those gaps. [How to Find Gaps in Your Literature Review and Fill Them Automatically](https://www.youtube.com/embed/ZH_0W6BSYvM) A **research gap** is a question the literature hasn't answered yet: a population nobody has studied, a method nobody has tried, or a result nobody has repeated. Gaps matter for two reasons. They tell you what's missing from your own reading, and they're often where new research projects come from. This lesson works in two steps: find the gaps in the papers you've saved, then search the whole index to fill them. Pick the path that matches how many papers you have. > [tab] A Collection (up to \~100 papers) > > A normal **Pro** search is enough for a project-sized Collection. > > **1. Ask for a gaps matrix** > > Open the Collection in **My Library**. It's already attached to the search bar. (From anywhere else, click **+** in the search bar to attach it.) > > ```text theme={null} > Create a research gaps matrix showing what topics or populations are > missing from this collection. > ``` > > Consensus searches within the Collection, reads the relevant papers, and returns a grid of topics against populations. Empty cells are the gaps. > > > [!TIP] > > > > Choose the axes yourself: "create a research gaps matrix that looks at \[variable A] by \[variable B]". > > **2. Fill the gaps from the full index** > > Remove the Collection from the search bar (click **×** on it, or choose **+** › **Search corpus** › **All**) so Consensus searches all 400M+ papers. > > ```text theme={null} > Find papers published in the last 5 years that fill those gaps. > ``` > > Consensus remembers the gaps from your last question and runs one search per gap. > > **3. Save what you find** > > Open a paper and use **Save** to add it to the Collection, or select several papers in **References** and save them at once. > [tab] Your whole Library (hundreds of papers) > > For large libraries, use a **Deep Review** so Consensus can work through more of your papers. > > [Find Research Gaps in Your Paper Library Using AI](https://www.youtube.com/embed/-3EKYsTy_Tg) > > **1. Ask a Deep question of your Library** > > From inside **My Library**, your Library is already attached. Click **Deep**, then ask: > > ```text theme={null} > What are the population gaps in the research I've collected on [topic]? > ``` > > Deep Review runs its own sub-questions across your papers only. It never touches the broader index in this step. The answer includes a visual map of where studies are missing. > > **2. Fill the gaps from the full index** > > Remove the Library from the search bar and ask: > > ```text theme={null} > Search the broader index to find papers that fill these gaps. Focus on > papers published in the last 10 years. > ``` > > ![The Deep step panel showing retrieved, eligible, and included counts and the searches it ran](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/deep-search-steps.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=5339244798bc7c1366b0156fa02c9c58) > > The Deep Review plan panel shows how many of your papers it screened, found eligible, and read. > [!NOTE] > > A gap in *your* Collection isn't always a gap in the literature. Step 2 checks the difference: if the full index has plenty of papers on it, you just hadn't found them yet. If it doesn't, you may have found a real research gap, which is often where new projects start. ## Try it on your topic Run a gaps matrix on your project Collection, then search the full index to fill one of the gaps. > [!TIP] > > **You should now have:** 2–3 gaps in your topic, and for each one, whether it's a gap in your reading or in the research itself. > [accordion] Go deeper: getting papers into a Collection > > Click **+ Add** in the top right of My Library to upload PDF, RIS, or BibTeX files (up to 20 at a time, 50 MB each, text-based PDFs only), paste DOIs, or import from Zotero. See [Library](/core-features/library) for the full walkthrough, including Zotero import. > [accordion] Go deeper: export the report > > Scroll to the bottom of an answer and click **Export**: **Copy text**, **Copy with citations** (APA, MLA, Chicago, Harvard, LaTeX, AMA/Numeric), or **Export as PDF**. # Read a thread Source: https://docs.consensus.app/learn/read-a-consensus-answer What every part of a search result means: the search trace, the written answer, the citations, and the reference list with its study types and citation counts. A Consensus result has two halves: the **answer** on the left and the **papers** behind it on the right. Learning to read both is what turns an AI summary into something you understand and can defend. ![A Consensus search about endometriosis pain treatments, with the written answer on the left and a list of reference papers on the right](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/results-page.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=fa7a9c1962d241a9e2ebebd57e9a4b32) The answer on the left, the References panel on the right. ## 1. The search trace The lines at the top of the answer (for example, **Pro · 3 steps**) show what Consensus actually did: what it searched for, how many papers matched, and how many it read. Click a step to see more. ![The steps at the top of a Consensus answer: Pro, 3 steps, a search for treatments reduce pain in endometriosis with 20.5M results, and Read abstracts and PDFs, 20](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/results-search-trace.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=6da182723eddb146bd58264472319c8e) One search that matched 20.5M papers, then 20 papers read from their abstracts and PDFs. Check this first. If Consensus searched for something different from what you meant, rephrase your question before reading the answer. ## 2. The answer The answer summarizes what the research says. Look for: * **Citations.** The grey tags in the text (like `Sun 2026`) link each statement to a paper. A small document icon with a tick next to a citation means Consensus read the full paper; without it, Consensus read only the abstract. Hover any citation to see the exact quote; see [Check a claim against the paper](/core-features/citation-grounding). * **Hedged language.** Words like "mixed", "limited", or "some studies" are the answer telling you the evidence isn't settled. Don't skip past them. * **Visuals**, depending on your question: tables and figures comparing studies, a **Claims & Evidence** table rating how well each claim is supported, a **Research Gaps** map, or a **timeline**. ### The Consensus Meter For a yes/no question, the answer opens with the **Consensus Meter**: how many of the papers lean **Yes**, **Possibly**, **Mixed**, or **No**. **N** is the number of papers counted. ![The Consensus Meter for the question Do treatments reduce pain in endometriosis, N = 20: Yes 90%, Possibly 5%, Mixed 5%, No 0%](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/results-consensus-meter.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=535a3315e81d233fede06da87508dad5) 18 of 20 papers say yes, treatments reduce endometriosis pain. A meter that's mostly one color means the research broadly agrees. A split meter is a sign to read [Find where studies disagree](/learn/where-studies-disagree). ## 3. The References panel Each paper in the panel shows: ![A reference card for a randomized trial in the Lancet: its key takeaway, a 6 Supporting Quotes chip and an RCT badge highlighted, a Rigorous Journal badge, then the year, citation count, authors, journal, and a PDF link](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/results-reference-card.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=38156f3f5bdf544253a4f19f8e6d1489) The supporting quotes and the study type are the two things to check first. | Item | What it tells you | | - | - | | **Key takeaway** | The paper's main finding in one sentence | | **Supporting quotes** | The passages from this paper the answer relies on. Click to read them. | | **Study type** badge | How the study was done (see below). Some papers also get badges like **Highly cited** or **Rigorous journal**. | | **Year** and **citations** | How recent it is, and how much attention it's had | | **Journal** | Where it was published | | **PDF** | The full text is available to open | Switch between the default list, a compact list, and a **Table** view with the buttons at the top of the panel. The table view adds columns like the paper's answer to your question and the population studied, so you can compare papers at a glance. ![The References panel in table view with Paper, Answer, and Population columns, and the table view button highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/results-table-view.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=cb03ac045077a6b649b1ba9013a09b8e) Table view: one row per paper, with columns you can compare. ## Study types, in plain English Consensus is able to extract a number of different study types from papers, which you can use in filters or see on the References tab. To search for only certain study types, see [Filter for strong, recent studies](/search-basics/find-high-quality-recent-papers#every-filter-explained). ![Consensus study type badges grouped into Experimental, Observational, Evidence Synthesis, and Other](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/study-types-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=f7a2f33710a5c9a0205f0f2d6de1a201) Not all study types are equally strong: 1. **Systematic review / meta-analysis**: combines many studies on the same question. A great place to start. 2. **RCT (randomized controlled trial)**: people are randomly assigned to a treatment or a comparison group. The best single study for cause and effect. 3. **Cohort, case-control, cross-sectional**: observational studies that track or compare groups without assigning treatments. Useful, but they show links, not proof of cause. 4. **Case study / case report**: one person or a small group. Good for rare situations and new ideas. The list above runs roughly from strongest to weakest evidence for a "does X cause Y?" question. Other types, like literature reviews and commentaries, summarize or discuss research without new data. See [Research terms in plain English](/learn/research-terms) for more. ## Try it on your topic Run a search on your topic, expand the search trace, and look through the References panel. > [!TIP] > > **You should now have:** 3 papers you'd read first, and a reason for each. For example: "the most recent meta-analysis", "the most cited RCT", "the only study in my population". # Research terms dictionary Source: https://docs.consensus.app/learn/research-terms A short glossary of the research words you'll see in Consensus and in papers, from corpus and preprint to meta-analysis and effect size. Keep this page open while you work through the guide. Terms are grouped by when you'll first meet them. ## Finding papers | Term | What it means | | - | - | | **Corpus** | The set of papers being searched. In Consensus: All, Medical, or My Library, chosen under **+** › **Search corpus**. | | **Peer review** | Other experts checked the paper before the journal published it. | | **Preprint** | A paper shared publicly before peer review. Useful for new results, but not yet checked. | | **Abstract** | The short summary at the top of a paper. | | **Full text** | The whole paper: introduction, methods, results, and discussion. | | **DOI** | A permanent ID for a paper, like `10.1000/182`. Paste it into Consensus to find that exact paper. | | **Citation count** | How many other papers cite this one. A rough sign of influence, not of quality. | | **Seminal paper** | A foundational paper that much of a field builds on. | | **Journal quartile (Q1–Q4)** | A journal's rank within its field. Q1 means the top 25%. | | **Retracted** | Withdrawn after publication, usually because of errors or misconduct. Don't cite it. | ## Study types | Term | What it means | | - | - | | **RCT (randomized controlled trial)** | People are randomly put into a treatment group or a comparison group. The strongest single study for cause and effect. | | **Cohort study** | Follows a group over time to see who develops an outcome. | | **Case-control study** | Compares people who have an outcome with people who don't, looking back for causes. | | **Cross-sectional study** | A snapshot of a group at one point in time. | | **Case study / case report** | A detailed look at one person, group, or event. | | **Literature review** | A summary of research on a topic, written by an expert. | | **Systematic review** | A review that follows a strict, repeatable method to find and judge every relevant study. | | **Meta-analysis** | Combines the numbers from many studies into one overall result. | ## Reading results | Term | What it means | | - | - | | **Population** | Who was studied, such as adults with type 2 diabetes. | | **Intervention** | The treatment or exposure being tested. | | **Outcome** | What was measured to judge the result. | | **Sample size (n)** | How many people or things were studied. Bigger is usually more reliable. | | **Effect size** | How big the difference or relationship is, not just whether one exists. | | **Statistically significant** | Unlikely to be due to chance alone. It doesn't mean the effect is large or important. | | **Correlation vs causation** | Two things happening together (correlation) doesn't prove one causes the other (causation). | | **Confounder** | A third factor that could explain a result, such as age or income. | | **Limitation** | A weakness the authors admit, usually near the end of the discussion. | ## Doing research | Term | What it means | | - | - | | **Literature review** (the task) | Reading and summarizing what's known about your topic, usually the first step of a project. | | **Research gap** | A question the research hasn't answered yet. | | **Backward citation tracing** | Reading the papers a paper cites, to see what it built on. | | **Forward citation tracing** | Reading the papers that cite a paper, to see what came after. | | **Snowballing** | Repeating citation tracing from paper to paper to build a reading list. | | **PICO** | A way to frame a question: Population, Intervention, Comparison, Outcome. | # Spot the main findings and themes Source: https://docs.consensus.app/learn/spot-main-findings Group the research on your topic into themes, rank them by how strong the evidence is, and put the main findings into your own words. Reading papers one at a time makes it hard to see the big picture. A quicker way in is to ask Consensus to sort the research into **themes** (groups of studies that take the same approach or reach similar conclusions), then check how strong each theme is. ## Step 1: Ask for the themes ```text theme={null} What are the main findings on treatments for endometriosis pain? Group them by approach. ``` Consensus searches, reads the strongest papers, and organizes the answer by theme, with the key claims and papers for each. > [!TIP] > > "Group by approach" works well for treatments and methods. You can also ask Consensus to group by population, by outcome, or by time period. ## Step 2: Rank them by strength of evidence ```text theme={null} Which of these approaches has the strongest evidence? Rank the key claims by how well they're supported. ``` Consensus weighs each theme by how many papers support it and how strong those studies are. Look for a **Claims & Evidence** table in the answer. ## Step 3: Check the top papers yourself For each theme, open the one or two papers Consensus leans on most. Hover the citations to see the quotes, and confirm the papers say what the answer claims. See [Check a claim against the paper](/core-features/citation-grounding). ## Step 4: Write it in your own words Close the answer and write one sentence per theme from memory. If you can't, re-read that section. This is the step that turns Consensus's summary into your own understanding. ## Try it on your topic Ask for the main findings on your topic grouped by approach, then rank them. > [!TIP] > > **You should now have:** 3–5 themes, each with a one-sentence summary in your own words and 1–2 key papers. > [accordion] Go deeper: themes in a fast-moving field > > If your field publishes a lot of new work, see [Catch up on a fast-moving field](/search-basics/catch-up-on-a-field) for grouping recent evidence and comparing the strongest papers side by side. # Find where studies disagree Source: https://docs.consensus.app/learn/where-studies-disagree Find the questions your field hasn't settled, understand why studies reach different results, and spot the limitations that matter. Research rarely agrees completely. Knowing where studies disagree, and why, is one of the most useful things you can bring to your first project meeting. It shows you understand the literature, not just its headlines. ## Step 1: Ask where the evidence conflicts ```text theme={null} Where do studies disagree about hormonal treatments for endometriosis pain, and why might they reach different results? ``` For a yes/no question, look for the **Consensus Meter**. It shows how many papers lean yes, no, or mixed, and colors the citations to match. ## Step 2: Put both sides next to each other ```text theme={null} Make a table of the studies that support this and the studies that challenge it, with the population, sample size, and study design for each. ``` Seeing the details side by side usually explains the disagreement. ## Why studies disagree When two studies conflict, check these first: | Difference | Example | | - | - | | **Population** | Teenagers vs adults, or one country vs another | | **Study design** | An RCT vs an observational study | | **Sample size** | 40 people vs 4,000 people | | **What was measured** | Self-reported pain vs a clinical scale | | **Time** | Results after 6 weeks vs after 2 years | Often both studies are right about their own setting, and the real finding is "it depends on X". ## Step 3: Read the limitations Every paper has a **limitations** section, usually near the end of the discussion. Ask Consensus to pull them out: ```text theme={null} What limitations do these studies report? ``` Then open one or two papers and read that section yourself. ## Try it on your topic Ask where studies on your topic disagree, then build the supporting-vs-challenging table. > [!TIP] > > **You should now have:** one open debate in your topic, with a paper on each side and a likely reason they disagree. > [accordion] Go deeper: compare more studies > > See [Compare studies side by side](/search-basics/build-a-comparison-table) for building a full comparison table with the exact columns you need. # Connect ChatGPT Source: https://docs.consensus.app/mcp-clients/chatgpt Connect the Consensus app in ChatGPT. **One-click:** open [Consensus in the ChatGPT app directory](https://chatgpt.com/plugins/plugin_asdk_app_6943e6f4a928819195962de16fb9ffe4) and click **Connect**. Or connect manually: **1. Open ChatGPT** Open [chatgpt.com](https://chatgpt.com) and sign in. **2. Find Consensus** Open **Apps** from the sidebar, search for **Consensus**, and click **Connect**. **3. Sign in** Sign in with your Consensus account when prompted. **4. Start searching** Ask a research question, or call it by name: "Use Consensus to find RCTs on…" Search results display in an interactive widget showing top papers with citations and direct links. > [!NOTE] > > Consensus also works with ChatGPT Deep Research. See [Using Consensus with ChatGPT Deep Research](/consensus-mcp#using-consensus-with-chatgpt-deep-research) below. ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Claude Code Source: https://docs.consensus.app/mcp-clients/claude-code Add the Consensus MCP server to Claude Code with one command. ```bash theme={null} claude mcp add --transport http consensus https://mcp.consensus.app/mcp ``` OAuth will open in your browser automatically on first use. You can also trigger it manually with the `/mcp` command inside Claude Code. ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Claude Desktop Source: https://docs.consensus.app/mcp-clients/claude-desktop Add Consensus to Claude on the web or in the Claude desktop app. **One-click:** open [Consensus in the Claude directory](https://claude.ai/directory/65247229-f0c7-49df-9044-fcbb8b3894c6) and click **Connect**. Or connect manually: **1. Open Settings** Open [claude.ai](https://claude.ai) or the Claude desktop app and go to **Settings**. **2. Browse connectors** Select **Connectors**, then **Browse connectors**. **3. Find Consensus** Search for **Consensus** and click **Connect**. **4. Sign in** Approve the sign-in with your Consensus account, or continue without one at reduced limits. **5. Start searching** Ask a research question in any chat. Consensus appears in the tools menu. ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Codex Source: https://docs.consensus.app/mcp-clients/codex Add the Consensus MCP server to OpenAI Codex. ```bash theme={null} codex mcp add consensus --url https://mcp.consensus.app/mcp codex mcp login consensus ``` Or add to `~/.codex/config.toml`: ```toml theme={null} [mcp_servers.consensus] url = "https://mcp.consensus.app/mcp" ``` ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Cursor Source: https://docs.consensus.app/mcp-clients/cursor Install the Consensus MCP server in Cursor. **One-click:** [Install in Cursor](https://cursor.com/en/install-mcp?name=consensus\&config=eyJ1cmwiOiJodHRwczovL21jcC5jb25zZW5zdXMuYXBwL21jcCJ9). Or add to `~/.cursor/mcp.json`: ```json theme={null} { "mcpServers": { "consensus": { "command": "npx", "args": ["-y", "mcp-remote@latest", "https://mcp.consensus.app/mcp"] } } } ``` ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Gemini Spark Source: https://docs.consensus.app/mcp-clients/gemini-spark Add Consensus as a custom app for Gemini Spark. Add Consensus as a custom app for Gemini Spark: **1. Open Connected Apps** Go to [gemini.google.com](https://gemini.google.com) and sign in with a personal Google Account, then click **Settings & help** → **Connected Apps**. **2. Add a custom app** Under **Custom apps for Spark**, click **Add a custom app** and paste the server URL: ```text theme={null} https://mcp.consensus.app/mcp ``` **3. Approve the connection** Click **Next** and follow the prompts to finish OAuth. If your client doesn't support Dynamic Client Registration, expand **Advanced features** and enter your credentials. **4. Start searching** Ask Gemini Spark a research question and it can call Consensus tools. > [!NOTE] > > Requires access to Gemini Spark, a personal Google Account, Keep Activity turned on, and is currently available in the US. Workspace and school accounts are not supported yet. ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Microsoft 365 Copilot Source: https://docs.consensus.app/mcp-clients/microsoft-365-copilot Connect Consensus in Microsoft 365 Copilot once your admin has added it. In Microsoft 365 Copilot, Consensus is a connector your admin sets up for the tenant. Once it's there, connect it yourself under **Sources**. [YouTube video player](https://www.youtube.com/embed/g2_WBlJupx4) **1. Open Copilot Chat settings** In Microsoft 365 Copilot, select the **gear icon** at the bottom of the left sidebar, then **Settings**, then **Sources**. **2. Connect Consensus** Find **Consensus for Copilot** and select **Connect**, then sign in with your Consensus account. **3. Ask a research question** Type a research question like 'What does the research say about creatine and cognitive function?' and Copilot will call Consensus to find relevant papers. > [!NOTE] > > Don't see Consensus under Sources? Your admin needs to add it first. Send them the [Microsoft 365 Copilot admin guide](/enterprise/microsoft-365-copilot). A [Consensus account](https://consensus.app/sign-up) is required and free to create. ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Microsoft 365 Copilot Studio Source: https://docs.consensus.app/mcp-clients/microsoft-365-copilot-studio Add Consensus as a custom MCP tool in Microsoft Copilot Studio. This allows you to add Consensus as a custom tool in Copilot studio so it can be used with agents or other automations you have have. This type of authentication will  **1. Open custom tools** Go to Copilot studio and find the 'tools' tab on the left hand side. ![Clean Shot 2026 08 31 At 17 17 16@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.17.16@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=b5f6a6010734943bb41b7d8351b2e952) **2. Create 'Create new tool'** ![Clean Shot 2026 08 31 At 17 19 50@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.19.50@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=cf3823beb65f9b462ad5c1f658e9b6ba) Click 'new tool' to add a custom tool **3. Click 'Model context protocol'** ![Clean Shot 2026 08 31 At 17 21 06@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.21.06@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=7bb99d2343cd51fa74f3ace9e886aabe) You can also use the API to do this. But to use the MCP, click 'model context protocol'. **4. Fill in the connector information** ![Clean Shot 2026 08 31 At 17 27 17@2x](https://mintcdn.com/consensus-f7e6ffce/Qlioe-rt4Y_Nb1Hf/images/CleanShot-2026-08-31-at-17.27.17@2x.png?fit=max\&auto=format\&n=Qlioe-rt4Y_Nb1Hf\&q=85\&s=557f2599cb9066c2bb71177b2bc4b402) **Server name** Consensus **Server Description** Consensus is the go-to MCP for academic research. Search 400M+ scholarly articles, including 350M peer-reviewed papers, from PubMed, Semantic Scholar, and ArXiv — filter by study type (RCTs, meta-analyses, systematic reviews), journal quality, sample size, and more. Get results with citations, study takeaways, and direct links, or run pre-built workflows for literature reviews, grant research, and evidence synthesis. **Server URL** [https://mcp.consensus.app/mcp](https://mcp.consensus.app/mcp) **Authentication** OAuth 2.0 **Type** Dynamic Discovery **5. Save and use!** Thats it! This custom tool can now be used and shared with other team members. Each team member will be able to connect to their own Consensus account when they use the tool. > [!NOTE] > > A [Consensus account](https://consensus.app/sign-up) is required. It is free to create. ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect other MCP clients Source: https://docs.consensus.app/mcp-clients/other-mcp-clients Connect any MCP client that supports remote servers. Any client that supports remote MCP servers (Streamable HTTP) works. Point it at `https://mcp.consensus.app/mcp`. If your client supports OAuth, authentication is handled automatically. **No OAuth?** Send your Consensus API key as a Bearer token instead. See [API key (Bearer token)](/consensus-mcp#connect-your-client) for the header and example configs. > [!NOTE] > > The `search` tool returns a text payload of ranked papers and citation instructions, written for AI agents. If your client needs structured paper data, use the [Consensus API](/api-get-started). ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Perplexity Source: https://docs.consensus.app/mcp-clients/perplexity Add Consensus from the Perplexity Computer connectors store. **1. Open the Consensus connector** Go to [Consensus in the Perplexity connectors store](https://www.perplexity.ai/computer/connectors?query=consensus\&connector=consensus). **2. Add the connector** Click **+ Add Connector**. **3. Sign in with Consensus** Complete the OAuth sign-in with your Consensus account. You need a [Consensus account](https://consensus.app/sign-up), which is free to create. **4. Confirm the connection** Click **Connect** on the OAuth confirmation screen. **5. Open Computer** Click **Computer**. **6. Ask a research question** Say "Use Consensus to..." followed by a scientific question, for example: "Use Consensus to identify the pros and cons of GLP-1s." ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect VS Code (GitHub Copilot) Source: https://docs.consensus.app/mcp-clients/vs-code-github-copilot Add the Consensus MCP server to VS Code with GitHub Copilot. Add to your MCP configuration: ```json theme={null} { "servers": { "consensus": { "url": "https://mcp.consensus.app/mcp" } } } ``` ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # Connect Windsurf Source: https://docs.consensus.app/mcp-clients/windsurf Add the Consensus MCP server to Windsurf. Add to your MCP configuration: ```json theme={null} { "mcpServers": { "consensus": { "serverUrl": "https://mcp.consensus.app/mcp" } } } ``` ## Next > [card] [Verify your connection](/consensus-mcp#verify-your-connection) > > Run a test search and fix common setup issues. > [card] [Plans and access](/mcp-plans-and-access) > > Papers per search, monthly calls, and rate limits for each plan. # MCP plans and access Source: https://docs.consensus.app/mcp-plans-and-access What each Consensus plan includes on the MCP server: papers per search, monthly calls, rate limits, result fields, complete text, and API key access. > [!NOTE] > > Your results depend on your Consensus plan, not your AI subscription. Claude Pro, ChatGPT Plus, and similar plans have no effect on Consensus results. > [!TIP] > > No account is needed to get started. Connect directly and try it out, then sign in for more results per search. ## What each plan includes | | No account | Free | Pro | Teams | Deep | Enterprise | | - | - | - | - | - | - | - | | Papers per search | 3 | 10 | Up to 300 | Up to 300 | Up to 750 | Up to 1,000 | | Included calls per month | No monthly limit | 30 | 500 | 500 | 2,000 | Custom | | Max calls with additional usage | — | 30 | 1,000 | 1,000 | 10,000 | Custom | | Rate limit | 5 requests/second, shared by all no-account users | 1 request/second | 3 requests/second | 3 requests/second | 3 requests/second | Custom | | Study type, takeaway, and DOI in results | No | No | Yes | Yes | Yes | Yes | | Pagination past the first page | No | No | Yes | Yes | Yes | Yes | | [Full-text excerpts](/api-full-text) | No | No | Yes | Yes | Yes | Yes | | [Complete text](/mcp-tools/get-full-text) of open-access papers | No | No | Yes | Yes | Yes | Yes | | Institutional full-text links\* | No | No | Yes | Yes | Yes | Yes | \* Returned on the first 50 results when you've selected your institution in Consensus settings. An organization key uses the institution of the person who created it. Create a free account at [consensus.app/sign-up](https://consensus.app/sign-up) or upgrade at [consensus.app/pricing](https://consensus.app/pricing). ## How calls are counted * **One call per 100 papers returned**, rounded up, with a minimum of one call per request. A search that returns 150 papers uses 2 calls. * **A [Thread](/mcp-tools/quickstart) is charged by the amount of work the agent does.** `mode` controls the effort: a `deep` run does more work than a `pro` run, so it costs more. A failed run costs nothing. While a Thread runs, its maximum possible cost counts as used. If that doesn't fit under your plan's limit, starting the Thread or adding a follow-up returns an error. * **One credit per full text returned.** [`get_full_text`](/mcp-tools/get-full-text) costs 1 credit for each paper whose text comes back, and a credit counts the same as a call. Papers that come back blocked or not found cost nothing. A paper charged to you in the last 24 hours normally comes back free, so continuing a long paper with its cursor usually costs nothing. * **MCP and API share one monthly pool.** Searches from your MCP connection and from your API key draw down the same allowance. * **Usage resets on the 1st of each month.** > [!NOTE] > > "Papers per search" is the per-request ceiling. Searches that don't set a page size still return the default of 20 papers, or your plan's cap if it's lower. ### Additional usage Pro, Teams, and Deep can keep making calls past the included amount, up to the plan's maximum, when the account has an active metered API subscription and additional usage is turned on. Additional usage is billed at **\$0.05 per call**. Free plans stop at 30 calls. ## API key access Any Consensus API key works on the MCP server. Include it as a Bearer token and the server applies the key owner's plan: a personal key gets your plan, and an organization key gets that organization's plan. Enterprise keys get custom limits. ```http theme={null} Authorization: Bearer YOUR_API_KEY ``` See [API key (Bearer token)](/consensus-mcp#connect-your-client) for example client configs. Setting Consensus up behind an AI gateway, in Microsoft 365 Copilot for a whole tenant, or in Copilot Studio? See the [Enterprise guides](/enterprise/overview). For enterprise access, [contact sales](https://consensus.app/home/contact/sales/). ## Related > [card] [Get started](/consensus-mcp) > > Connect Claude, ChatGPT, and other MCP clients in under two minutes. > [card] [API plans and access](/api-plans-and-access) > > The same plans on the REST API, with pagination and error details. # add_to_thread Source: https://docs.consensus.app/mcp-tools/add-to-thread The Consensus MCP add_to_thread tool: ask a follow-up in an existing Thread, keeping the earlier context. `add_to_thread` asks a follow-up in an existing Thread. The research agent reads the earlier turns and can reuse the papers it already found, so you can refer to them directly. Like [`create_thread`](/mcp-tools/create-thread), it returns right away, and your assistant polls [`get_thread`](/mcp-tools/get-thread) for the new answer. Use it for a new question you raise after seeing an answer: narrow the evidence, extend the scope, or act on the papers it cited, such as "save these to my Sleep collection". It isn't for splitting one request into steps. Put the whole request in the original `create_thread` brief instead. > [!NOTE] > > Requires a signed-in Consensus account, over OAuth or an [API key as a Bearer token](/consensus-mcp#connect-your-client). You can add only to your own Threads. ## Parameters | Parameter | Type | Required | Description | | - | - | - | - | | `thread_id` | string | Yes | The `thread_id` from `create_thread` or [`find_threads`](/mcp-tools/find-threads). | | `input_message` | string | Yes | The follow-up, up to 10,000 characters. Be specific and include every constraint. | | `mode` | string | No | `pro` (default) or `deep`. `deep` requires a paid plan. See [`create_thread`](/mcp-tools/create-thread#parameters). | | `filters` | object | No | Search filters for this follow-up. Same keys as [`create_thread`](/mcp-tools/create-thread#filters). They replace the agent's own filter choices, so include every constraint you want enforced. | | `attachments` | object | No | Papers and collections for this follow-up. Same shape as [`create_thread`](/mcp-tools/create-thread#attachments). | | `skill` | string | No | A named research workflow for this follow-up. Same options and rules as [`create_thread`](/mcp-tools/create-thread#skills): set `mode` to `deep` too. | | `title` | string | No | The Thread's title from `create_thread`, echoed back in the result. | ## Response The same dispatch result as [`create_thread`](/mcp-tools/create-thread#response), with the new `interaction_id` and `status: "running"`. Pass the `interaction_id` to `get_thread` to read only the new answer. ## Example prompts * "Of the trials you cited, keep only RCTs with more than 100 participants. Does the conclusion change?" * "Do the results differ for adults over 60?" * "Save the papers you cited to my Creatine collection." * "Put the three meta-analyses you cited into a new collection called Creatine reviews." ## Good to know * **Wait for the last run.** Call `add_to_thread` only once `get_thread` reports the Thread `idle` or `failed`. While the latest interaction is still `running`, it returns an error. Keep polling, then send the follow-up. * **Cost:** each follow-up is charged like a new run, by the amount of work the agent does. A failed run costs nothing. See [plans and access](/mcp-plans-and-access#how-calls-are-counted). * **Rate limit:** one `add_to_thread` call per second. # create_thread Source: https://docs.consensus.app/mcp-tools/create-thread The Consensus MCP create_thread tool: start a Thread that hands a research question to the Consensus research agent. `create_thread` starts a Thread: it hands a research question to the Consensus research agent and returns right away. The agent plans and runs many searches, reads full text, follows citations forward and backward, and writes a cited answer. Your assistant then polls [`get_thread`](/mcp-tools/get-thread) for the result. Use it for questions that need synthesis or analysis: what the evidence says, comparisons, mechanisms, literature reviews, and multi-step research tasks. For a single lookup of papers, use [`search`](/mcp-tools/search). The research agent can: * Search semantically and by keyword, and look up an exact paper by DOI, title, or author * Find papers by an author * Crawl citations forward and backward, and expand to influential, seminal, or similar papers * Read the full text of papers and quote them * Search attached papers and Library collections * List your Library collections and the papers in one (filtered by publication year or date added), create or rename a collection, and save papers it found into one. See [Library actions](#library-actions) It doesn't browse the open web and doesn't ask clarifying questions. > [!NOTE] > > Requires a signed-in Consensus account, over OAuth or an [API key as a Bearer token](/consensus-mcp#connect-your-client). ## Parameters | Parameter | Type | Required | Description | | - | - | - | - | | `input_message` | string | Yes | The complete research brief, up to 10,000 characters. Include the question, scope, population, timeframe, comparisons, constraints, and the output you want. | | `mode` | string | No | `pro` (default) reads about 20 papers and fits almost every question. `deep` reads about 50 papers, up to 100, and writes a structured literature review. `deep` requires a paid plan. | | `filters` | object | No | Search filters applied to every search the agent runs. See [Filters](#filters). | | `attachments` | object | No | Papers and Library collections to focus the research on. See [Attachments](#attachments). | | `skill` | string | No | A named research workflow to run. See [Skills](#skills). | ### Filters `filters` takes the same filters as [`search`](/mcp-tools/search#parameters). Every key is optional: * `year_min`, `year_max`, `month_min`, `month_max` (integers) * `study_types` (array, for example `["rct", "meta-analysis"]`) * `human`, `controlled`, `open_access`, `exclude_preprints`, `clinical_guideline` (booleans) * `sample_size_min`, `citation_min`, `duration_min`, `duration_max` (integers) * `sjr_max` (integer, 1–4): the journal quartile ceiling, where `1` is Q1, the highest. Set `1` for Q1 journals only, or `2` for Q1 and Q2. * `sjr_min` (integer, 1–4): the journal quartile floor. It excludes the better quartiles, so `2` keeps Q2–Q4. Don't use it to ask for top-tier journals. * `domain`, `country`, `publisher_name`, `journal_name` (strings) ```json theme={null} { "year_min": 2020, "study_types": ["rct", "meta-analysis"], "human": true, "sjr_max": 2 } ``` > [!WARNING] > > Filters you pass replace the agent's own filter choices for every search in the run. Include every constraint you want enforced, such as `study_types` alongside `sjr_max`. Or leave `filters` empty and state the constraints in `input_message`. ### Attachments Attach up to 50 papers and up to 10 Library collections. The agent reads attached papers before it answers. ```json theme={null} { "paper_attachments": [{ "paper_id": "" }], "collection_ids": [123] } ``` Paper IDs come from `citations[].paper_id` in an earlier [`get_thread`](/mcp-tools/get-thread#response) result. ### Skills Set `skill` to run a [Skill](/core-features/skills), a named research workflow, instead of the agent's open-ended plan. Set it only when you ask for one of these workflows, and set `mode` to `deep` too: | Skill | What it does | | - | - | | `literature-review` | Synthesizes the literature on a topic: searches from several angles, follows citations, and investigates contradictory evidence. | | `citation-snowballing` | Expands from known seed papers by following backward and forward citations. Attach or name the seed papers. | | `find-citations` | Finds research to support, qualify, or challenge the uncited claims in a draft. Put the draft in `input_message`. | | `verify-citations` | Checks whether the citations already in a draft support their claims, and suggests the smallest fix for weak ones. Put the cited draft in `input_message` or attach it. | > [!NOTE] > > `skill` appears only when skills are enabled for your account. It isn't available with an organization API key. A call with `skill` returns an error if `mode` isn't `deep`, or if the skill needs input the call didn't include. ## Response `create_thread` returns as soon as the agent starts, with `status: "running"`: ```json theme={null} { "thread_id": "", "interaction_id": "", "created_at": "2026-10-09T14:00:00Z", "title": "Creatine dose and working memory in healthy adults", "status": "running", "mode": "pro", "url": "https://consensus.app/search/creatine-dose-and-working-memory-in-healthy-adults//", "message_for_user": "Consensus research thread started: [View this in Consensus](https://consensus.app/search/…)", "poll_after_seconds": 5, "max_wait_seconds": 180, "next_steps": "Wait 5 seconds (use a sleep/timeout) between each get_thread(thread_id='') call. …" } ``` Trimmed. The payload also carries `render_instructions` that tell your assistant to share the Consensus link first and to wait for the final answer. | Field | Description | | - | - | | `thread_id` | Pass to [`get_thread`](/mcp-tools/get-thread) and [`add_to_thread`](/mcp-tools/add-to-thread). | | `interaction_id` | The run this call started. Pass to `get_thread` to read only this run. | | `url` | The Thread in Consensus. Empty for a Thread started with an organization API key, which has no Consensus page. | | `poll_after_seconds` | Wait this long between `get_thread` calls: 5 seconds for `pro`, 20 for `deep`. Polling faster doesn't speed up the agent. | | `max_wait_seconds` | When to stop polling: 180 for `pro`, 600 for `deep`. | ## Library actions The agent works in your [Library](/core-features/library) when the brief asks it to. There's no separate Library tool: put the request in `input_message`, and name collections the way they appear in Consensus. If two collections share a name, also name the parent collection. | Action | Example request | | - | - | | List collections, with their paper counts and your role on each | "List all of my Library collections and how many papers each has." | | List a collection's papers, filtered by publication year, date added, or a title, author, or journal match | "Which papers did I add to my Sleep collection last month?" | | Create a collection, at the top level or inside another one | "Create a collection called Magnesium inside my Sleep collection." | | Save papers into a collection | "Find 10 RCTs on magnesium and sleep quality and save them to my Sleep collection." | | Rename a collection | "Rename my Sleep collection to Sleep and Supplements." | * The agent saves papers without pausing for confirmation, and creates a collection only when you ask for one. An unknown collection name comes back as not found. * Removing papers, deleting a collection, and moving a collection need a confirmation step, so they're available only in the Consensus web app. * A listing stops at 100 collections. The rest are summarized as "and N more", with no way to page through them. * To focus a later Thread on a collection, pass its id as `attachments.collection_ids`. The id is the number in the collection's web URL, `consensus.app/library/{id}/`. * A Thread started with an organization API key has no Library, so it can't list or change collections. ## Example prompts * "Start a Consensus research thread: what dose of creatine improves working memory in healthy adults, and how strong is the evidence?" * "Use Consensus to do a deep literature review on GLP-1 receptor agonists and cardiovascular outcomes since 2015." * "Trace the citations from Cong et al. 2013 and summarize how CRISPR-Cas9 editing evolved." * "Find 10 RCTs on magnesium and sleep quality and save them to my Sleep collection." * "List all of my Consensus Library collections with their names and number of papers." ## Good to know * **One brief per request.** Put the whole request in one `create_thread` call, including ordered sub-tasks. Splitting it into a chain of Threads gives worse results and adds a full agent run per step. * **Cost:** charged by the amount of work the agent does, so a `deep` run costs more than a `pro` run. A failed run costs nothing. While the Thread runs, its maximum possible cost counts as used. If that doesn't fit under your plan's limit, the call returns an error. See [plans and access](/mcp-plans-and-access#how-calls-are-counted). * **Rate limit:** one `create_thread` call per second. # find_threads Source: https://docs.consensus.app/mcp-tools/find-threads The Consensus MCP find_threads tool: list your Threads and find an earlier one by title. `find_threads` lists your Threads, so your assistant can pick up earlier research. Pass `query` to match on title. Each result carries the `thread_id` to pass to [`get_thread`](/mcp-tools/get-thread) or [`add_to_thread`](/mcp-tools/add-to-thread), and a link to the Thread in Consensus. > [!NOTE] > > Requires a signed-in Consensus account, over OAuth or an [API key as a Bearer token](/consensus-mcp#connect-your-client). ## Parameters | Parameter | Type | Required | Description | | - | - | - | - | | `query` | string | No | Match Threads whose title contains this text. Omit to list all your Threads. | | `limit` | integer | No | Maximum Threads to return, from 1 to 100. Defaults to 20. | | `offset` | integer | No | Number of Threads to skip, for paging. Defaults to 0. | ## Response ```json theme={null} { "threads": [ { "thread_id": "", "title": "Creatine dose and working memory in healthy adults", "status": "idle", "preview": "What dose of creatine improves working memory in healthy adults…", "interaction_count": 2, "last_activity_at": "2026-10-09T14:05:00Z", "url": "https://consensus.app/search/creatine-dose-and-working-memory-in-healthy-adults//" } ], "has_more": false } ``` Illustrative. `status` is `running`, `idle`, or `failed`. When `has_more` is `true`, call again with `offset` raised by `limit` to get the next page. ## Example prompts * "Find my Consensus thread on creatine and pick up where we left off." * "List my recent Consensus research threads." ## Good to know * **Rate limit:** one `find_threads` call per second. # get_full_text Source: https://docs.consensus.app/mcp-tools/get-full-text The Consensus MCP get_full_text tool: read the complete text of open-access papers as Markdown, by the ids from search results or Threads, or by DOI. `get_full_text` reads the complete text of open-access papers, as Markdown. [`search`](/mcp-tools/search) returns abstracts and short excerpts. When your assistant needs a paper's methods, results tables, or full argument, it passes the paper's id from search results or a Thread's citations to `get_full_text`. It works the same way in every MCP client, ChatGPT included. ## How it works **1. Search or run a Thread** Your assistant calls `search`, or reads a Thread with [`get_thread`](/mcp-tools/get-thread). Each search result has an `id:` line, marked `full text available` when the paper is open access and `get_full_text` is available to you. Each paper a Thread cites has a `paper_id` in its `citations`. **2. Read** Your assistant calls `get_full_text` with up to 3 of those ids, or DOIs. Each paper comes back as Markdown, up to 40,000 characters at a time. **3. Continue a long paper** A paper longer than 40,000 characters ends with a `Continue: get_full_text(cursor="...")` line. Calling it returns the next part, at no extra cost. If the line ends with `· may cost 1 credit`, Consensus couldn't promise a free re-read this time, so the next part may be charged again. ## Parameters Pass `paper_ids`, `dois`, or both, up to 3 combined. To read the next part of a paper, pass `cursor` on its own instead. | Parameter | Type | Required | Description | | - | - | - | - | | `paper_ids` | string array | No | Paper ids: the `id:` shown under each `search` result, or a `paper_id` from a Thread's `citations`. Up to 3 combined with `dois`. | | `dois` | string array | No | DOIs, with or without a `https://doi.org/` prefix. Up to 3 combined with `paper_ids`. | | `cursor` | string | No | The cursor from a `Continue:` line, to read that paper's next part. Pass it alone, without `paper_ids` or `dois`. | ## Response `get_full_text` returns text: each paper you asked for, in order, then the credits the call used. ```text theme={null} 3f1c... · chars 0 to 39,912 of 61,234 ## Introduction Creatine is ... ## Methods ... Continue: get_full_text(cursor="") 8b2d... · blocked: not open access, retracted, or no open-access copy held https://consensus.app/papers/details/8b2d.../?utm_source=claude_code Charged 1 credit. ``` Illustrative and trimmed. Each paper starts with a header line: its id (and its DOI, when you passed one), then which characters of its Markdown this part covers. A paper that doesn't come back says why instead: | Header ends with | Meaning | Charged | | - | - | - | | `chars to of ` | The paper is open access. This part of its Markdown follows. | 1 credit, once a day | | `blocked: not open access, retracted, or no open-access copy held` | The paper isn't open access, it was retracted, or its copy in Consensus isn't its open-access copy, so no text comes back. Its Consensus link follows. | No | | `no full text: open access, but Consensus holds no text for it` | The paper is open access, but there's no text to return. Its Consensus link follows. | No | | `not found: no Consensus paper has this id` (or `DOI`) | No Consensus paper matches. | No | Which papers count as open access is the same rule the API uses: gold, diamond, green, and hybrid papers whose text came from an open source such as arXiv or PubMed Central, never from a publisher feed. See [which papers have complete text](/api-full-text#which-papers-have-complete-text). ## Example prompts * "Find open-access RCTs on creatine and working memory, then read the methods of the two largest." * "Read the full text of 10.2903/j.efsa.2024.9100 and summarize how the panel weighed the evidence." * "Read the two papers that Thread cites most and check whether their dosing protocols match." * "Search Consensus for meta-analyses on mindfulness for anxiety and compare the inclusion criteria of the open-access ones." ## Good to know * **Cost:** 1 credit per paper returned, from the same monthly allowance as `search` and the API. Blocked and not-found papers cost nothing. A paper charged to you in the last 24 hours normally comes back free, so continuing with a cursor or reading a paper again the same day usually costs nothing. When a `Continue:` line ends with `· may cost 1 credit`, the next part may be charged. See [plans and access](/mcp-plans-and-access#how-calls-are-counted). * **Cursors expire:** a cursor works for about a day. After that, call `get_full_text` with the paper's id again, which may cost 1 credit. * **Plans:** requires a paid plan. On a Free plan or without an account, your assistant doesn't see the tool. Upgrade to a paid plan to use it. * **Rate limit:** 3 `get_full_text` calls per second. * **Same as the API:** `get_full_text` returns the same text as [`GET /v1/papers/full_text`](/api-full-text#complete-text), which takes up to 20 papers per request. # get_thread Source: https://docs.consensus.app/mcp-tools/get-thread The Consensus MCP get_thread tool: poll a Thread and read the research agent's cited answer when it's ready. `get_thread` polls a Thread. While the agent works, `status` is `running`. When it's `idle`, the result carries the agent's answer: markdown with inline citations, the grounding quotes, the cited papers, and a ready-to-write document. Your assistant calls it after [`create_thread`](/mcp-tools/create-thread) or [`add_to_thread`](/mcp-tools/add-to-thread), about every 5 seconds for `pro` and every 20 seconds for `deep`. A `pro` run usually finishes in under a minute. A `deep` run takes about 2 to 4 minutes, and occasionally up to about 10. Polling faster doesn't speed up the agent. > [!NOTE] > > Requires a signed-in Consensus account, over OAuth or an [API key as a Bearer token](/consensus-mcp#connect-your-client). You can read only your own Threads. ## Parameters | Parameter | Type | Required | Description | | - | - | - | - | | `thread_id` | string | Yes | The `thread_id` from `create_thread` or [`find_threads`](/mcp-tools/find-threads). | | `interaction_id` | string | No | Return only this interaction, for example the run you just started. Omit to return every interaction in the Thread. | | `include_trace` | boolean | No | Set to `true` to add `agent_trace`, a plain-English list of the steps the agent took (the searches it ran, the papers it read). Off by default to keep polls light. | ## Response ```json theme={null} { "thread_id": "", "title": "Creatine dose and working memory in healthy adults", "filename": "Creatine dose and working memory in healthy adults.md", "status": "idle", "url": "https://consensus.app/search/creatine-dose-and-working-memory-in-healthy-adults//", "message_for_user": "Your Consensus research thread is ready. [View this in Consensus](https://consensus.app/search/…)", "interactions": [ { "interaction_id": "", "input_message": "What dose of creatine improves working memory in healthy adults…", "status": "idle", "mode": "pro", "response": "Short-term loading at about 20 g/day improved working memory in some trials [[1.1]](https://consensus.app/papers/…) …", "document": "# Creatine dose and working memory in healthy adults\n\n… \n\n## References\n…", "citations": [ { "n": 1, "paper_id": "", "title": "Creatine and improvement in cognitive function: Evaluation of a health claim …", "year": 2024, "journal": "EFSA Journal", "url": "https://consensus.app/papers/…", "snippets": [ { "marker": "1.1", "quote": "…", "section": "Weighing of the evidence" } ] } ], "stats": { "retrieved": 120, "eligible": 40, "included": 18 }, "step_count": 12 } ] } ``` Illustrative and trimmed. The payload also carries `render_instructions` and a `schema_note` for your assistant. | Field | Description | | - | - | | `status` | `running`, `idle` (answer ready), or `failed`. Matches the latest interaction. | | `url` | The Thread in Consensus, where the interactive consensus meter and citation graph live. Empty for a Thread started with an organization API key. | | `filename` | A file-safe name your assistant can save `document` under. | | `interactions[].skill` | The [Skill](/core-features/skills) the interaction ran, such as `literature-review` or `verify-citations`: the one requested, otherwise the one the agent chose itself. Absent when none applies. | | `interactions[].response` | The answer in markdown. A marker like `[1.1]` cites quote 1 from paper 1, and a bare `[1]` cites the paper without a quote. Each marker links to its paper. | | `interactions[].document` | A complete markdown document: `response`, a Consensus disclaimer, and a References section. Present only when the interaction is `idle`. | | `interactions[].citations` | The cited papers, numbered by first citation in `response`. Each `snippets[].marker` matches a marker in the text. On a paid plan, if you've selected your institution in Consensus settings, a citation can include an `institutional_access_url` that opens the full text through your library, and its entry in the `document` References links to it. Pass a citation's `paper_id` to [`get_full_text`](/mcp-tools/get-full-text) to read that paper in full. | | `interactions[].stats` | Papers `retrieved`, `eligible`, and `included` in the answer. | | `interactions[].step_count` | How many steps the agent took. Matches the steps listed in `agent_trace`. | | `interactions[].agent_trace` | The agent's steps as markdown bullets, then the paper funnel. Present only with `include_trace=true`. Written for display, so don't parse it: the wording can change. | Only an `idle` interaction carries the full `response` and `citations`. Evidence-strength scores, top authors, the results timeline, the research-gap matrix, and the search funnel come back as text and tables inside `response` and `document`. ## How assistants show the result `get_thread` tells your assistant to: 1. Share the **View this in Consensus** link first. 2. Write `document` to the file named in `filename`, or render it in an artifact, canvas, or side panel, instead of pasting it into chat. 3. Keep every citation marker exactly as written. 4. Reply in chat with a 2–3 sentence summary and the Consensus link. ## Good to know * **Rate limit:** up to three `get_thread` calls per second. * **Wrong interaction:** an `interaction_id` from another Thread returns an error. Poll with the `thread_id` alone to get the latest interaction. # Quick start: Search & Threads Source: https://docs.consensus.app/mcp-tools/quickstart The two ways an assistant uses Consensus MCP: search for ranked papers in one call, or hand a research question to the Consensus research agent with Threads. Consensus MCP gives your assistant two ways to work with the literature: * **Search:** the `search` tool returns ranked papers in one call. Your assistant reads them and writes the answer. * **Threads:** the thread tools hand the whole research question to the Consensus research agent. The agent searches, reads the papers, and writes a cited answer. Your assistant starts the Thread, waits for it, and shows the result. Ask your question in plain language and your assistant picks the right tool for it. To make it use a specific one, say so in your prompt: for example, "Search Consensus for…" for search, or "Start a Consensus research thread on…" for a Thread. ## Search vs Threads | | Search | Threads | | - | - | - | | **Tools** | [`search`](/mcp-tools/search), then [`get_full_text`](/mcp-tools/get-full-text) to read open-access papers in full | [`create_thread`](/mcp-tools/create-thread), [`get_thread`](/mcp-tools/get-thread), [`add_to_thread`](/mcp-tools/add-to-thread), [`find_threads`](/mcp-tools/find-threads) | | **Your assistant gets back** | Ranked papers with abstracts and metadata | A written analysis with inline citations, grounding quotes, and the cited papers | | **Functionality** | Semantic paper search with filters | Search, citation crawls, author search, finding specific papers, reading full papers, searching attached papers and collections, and saving papers to your Library | | **Steps per request** | One | Many | | **Timing** | Seconds | Minutes. The assistant polls until the answer is ready | | **Follow-ups** | Each call stands alone | Follow-ups build on the earlier answers in the same Thread | | **Account** | Works without an account at reduced limits | Requires a signed-in Consensus account or API key | Need a quick lookup, or papers for your assistant to reason over? Use search. Need a researched, cited answer to a question that takes several steps? Use a Thread. ## Search Your assistant calls `search` with a query and any filters you asked for, then answers from the ranked papers it gets back. Try asking: * "Find RCTs and meta-analyses since 2020 on cognitive behavioral therapy for anxiety" * "Search for human studies on gut microbiome and mental health with at least 100 participants" * "Recent research on large language model hallucination from Q1 journals" See the [`search` tool reference](/mcp-tools/search) for every filter. ### Read whole papers When the abstract and excerpts aren't enough, your assistant can read a paper's complete text. Each search result shows an `id:` line, marked `full text available` when the paper is open access and `get_full_text` is available to you. Your assistant passes up to 3 of those ids, a Thread citation's `paper_id`, or DOIs, to [`get_full_text`](/mcp-tools/get-full-text) and gets each paper back as Markdown. It costs 1 credit per paper returned and needs a paid plan. * "Read the methods section of the largest RCT in those results." * "Get the full text of 10.2903/j.efsa.2024.9100 and list every dose it reports." ## Threads Threads are asynchronous. A `pro` run usually finishes in under a minute, and a `deep` run takes a few minutes. **1. Start a Thread** Your assistant calls [`create_thread`](/mcp-tools/create-thread) with the complete research brief. It returns a `thread_id` and a link to the Thread in Consensus right away, with `status: "running"`. **2. Wait for the answer** Your assistant calls [`get_thread`](/mcp-tools/get-thread) about every 5 seconds for `pro` and every 20 seconds for `deep`, until `status` is `idle` (answer ready) or `failed`. A `pro` run usually finishes in under a minute. A `deep` run takes about 2 to 4 minutes, and occasionally up to about 10. **3. Show the result** When the Thread is `idle`, `get_thread` returns the analysis as a ready-to-write markdown document with a References section. Most assistants write it to a file, artifact, or canvas and reply with a short summary plus the Consensus link. **4. Ask a follow-up** [`add_to_thread`](/mcp-tools/add-to-thread) continues the same Thread. The agent keeps the earlier context, so you don't need to restate it. To pick up an earlier Thread, ask your assistant to find it. [`find_threads`](/mcp-tools/find-threads) lists your Threads and can match on title. ### What you can ask a Thread Each of these takes several research steps, which a single search can't do. > [card] Crawl citations > > "Start from Cong et al. 2013, the first report of CRISPR-Cas9 editing in mammalian cells. Trace the work it built on and the most-cited papers that followed, and summarize how the method evolved." > [card] Research an author > > "Summarize Walter Willett's research on dietary fat and heart disease, and how his conclusions changed over time." > [card] Run deep research > > "Do a deep literature review on GLP-1 receptor agonists and cardiovascular outcomes. Cover mechanisms, the major trials, safety signals, and open questions." > [card] Save papers to your Library > > "Find 10 RCTs on magnesium and sleep quality and save them to my Sleep collection." > [card] Build on an earlier answer > > "Of the trials you cited, keep only RCTs with more than 100 participants. Does the conclusion change?" ## Good to know * **Account:** the thread tools appear only for a signed-in Consensus account, over OAuth or an [API key as a Bearer token](/consensus-mcp#connect-your-client). Without an account, your assistant sees only `search`. * **Deep mode:** `deep` reads about 50 papers and writes a longer, structured report. It requires a paid plan, and your assistant uses it only when you ask for a deep or comprehensive review. * **Cost:** search costs one call per 100 papers returned, and `get_full_text` 1 credit per paper returned. A Thread draws from the same monthly calls and is charged by the amount of work the agent does, so a `deep` run costs more than a `pro` run. A failed run costs nothing. See [plans and access](/mcp-plans-and-access#how-calls-are-counted). * **In Consensus:** a Thread started from your account also appears in Consensus, with the interactive consensus meter and citation graph. Open it from the `url` link. * **Same as the API:** the thread tools run the same research agent as the [Threads API](/api-quickstart-threads). ## Tool reference > [card] [search](/mcp-tools/search) > > Find ranked papers for a query in one call. > [card] [get_full_text](/mcp-tools/get-full-text) > > Read the complete text of open-access papers. > [card] [create_thread](/mcp-tools/create-thread) > > Start a Thread with the research agent. > [card] [get_thread](/mcp-tools/get-thread) > > Poll a Thread and read the cited answer. > [card] [add_to_thread](/mcp-tools/add-to-thread) > > Ask a follow-up in an existing Thread. > [card] [find_threads](/mcp-tools/find-threads) > > List your Threads and find an earlier one. # search Source: https://docs.consensus.app/mcp-tools/search The Consensus MCP search tool: find ranked, peer-reviewed papers for a query in one call, with academic filters. `search` finds peer-reviewed papers for a query and returns them ranked, in one call. Your assistant reads the papers and writes the answer. It's the right tool for lookups and for questions your assistant can answer from a list of papers. For a question that needs several research steps and a written, cited answer, use [`create_thread`](/mcp-tools/create-thread) instead. See [Search vs Threads](/mcp-tools/quickstart#search-vs-threads). `search` works without an account at reduced limits. ChatGPT Deep Research may call it several times with different queries and filters while it builds a report. ## Parameters Every parameter except `query` is optional. | Parameter | Type | Required | Description | | - | - | - | - | | `query` | string | Yes | The research question or topic, up to 500 characters. Use specific academic terminology for best results. | | `year_min` | integer | No | Exclude papers published before this year. | | `year_max` | integer | No | Exclude papers published after this year. | | `month_min` | integer | No | Minimum publication month (1–12). Use with `year_min`. | | `month_max` | integer | No | Maximum publication month (1–12). Use with `year_max`. | | `study_types` | string array | No | Filter by study design. Values: `rct`, `meta-analysis`, `systematic review`, `literature review`, `case report`, `non-rct experimental`, `non-rct observational study`, `non-rct in vitro`, `animal`. | | `sjr_max` | integer | No | Maximum journal quartile (SCImago Journal Rank). `1` is Q1 (highest), `4` is Q4. Set to `1` for Q1 journals only. | | `sjr_min` | integer | No | Minimum journal quartile. Excludes better (lower-number) quartiles. For example, set to `2` to keep Q2–Q4. | | `human` | boolean | No | Set to `true` to include only studies involving human subjects. | | `controlled` | boolean | No | Set to `true` to include only controlled studies. | | `sample_size_min` | integer | No | Exclude studies with fewer participants than this number. | | `citation_min` | integer | No | Exclude papers with fewer citations than this number. | | `medical_mode` | boolean | No | Set to `true` for clinical, medical, or evidence-based medicine questions. Limits results to top medical journals and clinical guidelines. | | `exclude_preprints` | boolean | No | Set to `true` to return peer-reviewed papers only. | | `open_access` | boolean | No | Set to `true` to include only open-access papers. | | `duration_min` | integer | No | Minimum study duration in days. Useful for longitudinal or long-term studies. | | `duration_max` | integer | No | Maximum study duration in days. | | `country` | string | No | Comma-separated ISO 3166-1 alpha-2 country codes (for example `us,gb`) to limit to those countries of study. Unknown codes are ignored. | | `publisher_name` | string | No | Comma-separated publisher names, for example `Elsevier`. | | `journal_name` | string | No | Preferred journal, for example `Nature`. Boosts matching papers without excluding others. | | `domain` | string | No | Comma-separated academic field codes, for example `med` or `bio,cs`. | | `page` | integer | No | Zero-indexed result page, up to `49`. Pages after the first need a paid plan. | | `page_size` | integer | No | Results per page. Defaults to 20, capped by your plan. | | `include_full_text_chunks` | boolean | No | Query-relevant [full-text excerpts](/api-full-text) from open-access papers, up to 3 per paper. **Default:** on for paid plans (Pro, Deep, Teams, Enterprise), off on Free and without an account. Set `false` to leave excerpts out. Setting `true` on Free or without an account returns an error. | ## Response `search` returns a text payload written for AI agents: the ranked papers (title, link, authors, year, citation count, journal, paper id, abstract, and on paid plans full-text excerpts), plus instructions to cite each paper inline by its number. ```text theme={null} Found 20 papers, showing top 20. [1] [Caffeine and endurance performance: a meta-analysis](https://consensus.app/papers/details/3f1c.../?utm_source=claude_code) (Smith et al., 2021, 154 citations, Sports Medicine, DOI: 10.1007/s40279-021-01470-1) id: 3f1c... · full text available Access through your institution: Abstract text... Full-text excerpts: - Section: Results | Excerpt text... - Section: Discussion | Excerpt text... [2] ... IMPORTANT INSTRUCTIONS: When discussing these findings, you MUST cite papers inline using their numbered references, e.g. [1], [2]. ... ``` Paid plans also get DOIs and full-text excerpts from open-access papers: up to 3 per paper for the first 5 eligible papers, and 12,000 characters per search. When a search reaches that limit, the excerpt at the limit is cut short with `...` and later papers show none. In ChatGPT, results appear in the Consensus widget. Each paper's `id:` line carries the id to pass to [`get_full_text`](/mcp-tools/get-full-text), which reads the paper's complete text. Papers marked `full text available` are open access, so their complete text can be returned. The mark only appears when `get_full_text` is available to you: on a Free plan, or before the tool reaches your account, id lines carry no mark. ChatGPT gets the same in each paper's `paper_id` and `full_text_available` fields. On a paid plan, if you've selected your institution in Consensus settings, a paper in the first 50 results can carry an **Access through your institution** link that opens the full text through your library. ChatGPT gets it as the paper's `institutional_access_url`. > [!NOTE] > > Need structured JSON? Use the [Consensus API](/api-quickstart-search), which shares your monthly allowance with MCP. ## Example prompts * "What does the research say about the effectiveness of remote work on productivity?" * "Find RCTs and meta-analyses since 2020 on cognitive behavioral therapy for anxiety" * "Search for high-quality human studies on gut microbiome and mental health with at least 100 participants" * "Recent research on large language model hallucination from top-tier journals" ## Good to know * **Cost:** one call per 100 papers returned, rounded up, with a minimum of one call per search. See [plans and access](/mcp-plans-and-access#how-calls-are-counted). * **Papers per search:** your plan sets the maximum `page_size`, from 3 without an account up to 1,000 on Enterprise. * **Rate limits:** set by your plan. See [plans and access](/mcp-plans-and-access#what-each-plan-includes). # What you can build with the Consensus MCP Source: https://docs.consensus.app/mcp-use-cases # Connect two fields Source: https://docs.consensus.app/search-basics/bridge-two-fields Bridge two disciplines in one thread: establish the limits in your field, find where they show up in another, map the gaps, and find collaborators. [How to Connect a Method From One Research Field to Applications in Another](https://www.youtube.com/embed/AIB1m1LzT2I) Some of the best research comes from porting a method between disciplines, but first you have to learn what keywords the other field uses. Consensus figures out the connections for you. The example: a physicist working on ultrasound in heterogeneous materials looks for applications in medical imaging. **You'll learn how to:** 1. Lay the foundation of a thread in your source field 2. Run a cross-domain search into a target field 3. Generate a research gaps matrix and find potential collaborators ## Step 1: Establish the source field Open a thread on the core topic so Consensus has the foundation for the follow-ups. ```text theme={null} Core limitations of ultrasound in heterogeneous materials. ``` Research Agent reads recent key papers and summarizes the core limitations. ## Step 2: Cross into the target field ```text theme={null} Find the strongest medical imaging papers where tissue heterogeneity is a main challenge for ultrasound. ``` Consensus works out the keywords the medical imaging literature uses, runs three parallel searches with different terms, and outlines the core areas and critical papers at the intersection. ## Step 3: Map the gaps ```text theme={null} What research is missing in this field? Can you create a research gaps matrix to show me? ``` Consensus loads the thread, runs additional searches across the subfields, and returns a topic × context matrix. In the example, in vivo, 3D, and real-time scenarios are poorly covered, and non-linear propagation and reverberation are mostly studied in simulation rather than patients. ## Step 4: Find collaborators ```text theme={null} Who are co-authors that I could collaborate with here on non-linear and reverberation-aware correction in vivo? ``` Consensus searches deeper on that specific topic and lists active researchers grouped by institution and sub-area. > [!NOTE] > > The video description mentions convolutional neural networks and medical imaging; the recorded demo uses ultrasound physics and tissue heterogeneity. ## Threads keep the whole investigation together Each follow-up runs inside the same thread, so Consensus carries forward what it learned. Type in **Ask a follow up…** at the bottom of the page; you can apply filters to a follow-up too. ![The Ask a follow up box beneath a Pro answer](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/threads-02.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=a5cd266a0bdbb2470c94d8db923604a5) * Start a fresh thread with **+ New Thread** or **Cmd/Ctrl + K**. * Click the thread title to rename it, jump between queries with the target icon, or delete a query with the three-dot menu. * Click **Share**, set visibility to **Public**, and **Copy thread link** to send the whole investigation to a collaborator. ![The Thread details dialog with the name field and query list](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/threads-05.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=77db89613166f618c14d84d5a57ff4d5) Thread details: rename the thread and manage its queries. ![The Share menu with Private and Public visibility options and Copy thread link](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/threads-03.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ee07cc1f0538e20323fbd127f71e8dee) ## Related tutorials > [card] [Look up a researcher's work](/search-basics/research-an-author) > [card] [Grow a reading list from one paper](/search-basics/snowball-reading-list) # Compare studies side by side Source: https://docs.consensus.app/search-basics/build-a-comparison-table Compare interventions side by side with the same metrics. Run parallel searches, then extract sample size, session count, follow-up duration, and effect size into one table. [How to Build a Comparison Table of Research Studies](https://www.youtube.com/embed/WTe9XsY1lNs) Comparing interventions across studies usually means hours of pulling values out of PDFs. This tutorial builds a side-by-side table in two searches, using RCTs on CBT, IPT, and digital interventions for adolescent depression. **You'll learn how to:** 1. Run parallel searches across multiple interventions in one query 2. Extract sample size, session count, follow-up duration, and effect sizes into a structured table 3. Use the table to spot the strongest intervention and evidence gaps at a glance ## Step 1: Search several interventions at once ```text theme={null} Find RCTs on CBT, IPT, and digital interventions for adolescent depression. ``` Research Agent runs three searches in parallel, one per intervention, each with a **randomized controlled trial** filter applied. It finds over 300,000 papers, deduplicates them, and shortlists the 20 most relevant. The answer already includes a table of populations, interventions, and key findings. > [!TIP] > > Click the **filter** button in the search bar to see every study-type filter: RCT, case study, in vivo, in vitro, human studies only, and more. ## Step 2: Ask for the exact columns you need ```text theme={null} Build a comparison table with columns: intervention type, sample size, session count, follow-up duration, and effect size on depression symptoms. ``` Consensus loads the papers from the previous thread, reads the full text where available (or the abstract), narrows to 10 studies, and returns the exact table with a link to each paper. A document icon with a tick next to a citation means the full text was used. Click the **PDF** button to read the study inside Consensus. ## Step 3: Ask a paper about itself Attach a paper to the search bar and ask a follow-up: ```text theme={null} What gaps did the authors mention? ``` Consensus reads the entire full text. Hover a citation in the answer to see the quote it came from. To read it in context, click the paper's **Supporting Quotes**, open the **Attachment** tab to see the PDF, and pick a quote from the menu at the bottom to highlight it. > [!NOTE] > > If Consensus doesn't have the full text, attach the PDF yourself or [import your Zotero library](/core-features/library). Uploads are private to you. ## Table view: the built-in comparison You don't always need a prompt. Open **References** on the right and click the **Table** view button to extract study attributes side by side for every paper in the results: key takeaway, population, methods, results, outcomes, sample size, study count, duration, and country. Fields Consensus can't extract are left blank. ![Search results in table view with Paper, Answer, and Population columns](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/results-view-03.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=30c01abd7747743f621a731d6cecb58f) Table view, toggled from the buttons at the top right of the results. | Plan | Papers per query in table view | | - | - | | Free | 3 | | Pro | 20 | | Deep | 50 | ## Attach papers straight from the results Tick the boxes next to specific papers in the results list and ask your question in the search bar at the bottom. Only those papers are used, and Consensus reads their full text where available. ![Search results with two papers checked and attached to the follow-up search bar](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/chat-with-full-text-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=a7f2f4f34725e380c8e449dccdc793b4) Selected papers appear as chips above the search bar. ## Export the table or the papers * **The answer** — scroll to the bottom and click **Export** for **Copy text**, **Copy with citations** (APA, MLA, Chicago, Harvard, LaTeX, AMA/Numeric), or **Export as PDF**. Tables survive a paste into Word, Docs, Notion, or Obsidian. * **The papers** — click the download arrow at the top of the results for **CSV**, **RIS** (EndNote, Zotero, Mendeley), or a **Bibliography**. ![The export menu on the results panel with CSV, RIS, and Bibliography options](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/pro-messages-05.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=34fefbf55ef6088db4d48d2e849557c6) ## Related tutorials > [card] [Filter for strong, recent studies](/search-basics/find-high-quality-recent-papers) > [card] [Check a manuscript's citations](/search-basics/manuscript-citation-check) # Catch up on a fast-moving field Source: https://docs.consensus.app/search-basics/catch-up-on-a-field Fifty new papers a week and no time to read them. Build a what's-new-and-what's-strongest digest grouped by approach, then drill into the winning papers. [How to Catch Up on a Fast Moving Research Field in 5 Minutes](https://www.youtube.com/embed/ktINOSj8Kv0) You don't need to read every new paper. You need to know what's new, organized by approach, and which approaches actually have evidence behind them. The example: techniques for reducing hallucinations in large language models since 2024. **You'll learn how to:** 1. Run a recency-filtered search with results grouped by approach 2. Rank approaches by empirical support across studies 3. Drill into the strongest papers and jump straight to the passage in the PDF ## Step 1: Recent evidence, grouped by approach ```text theme={null} What is the most compelling evidence for methods of reducing hallucinations in large language models published since 2024? Group them by approach. ``` Research Agent runs parallel searches, all with a **2024+** filter, narrows over 2 million candidates to 185 relevant papers, and shortlists 20. The answer groups methods into families with key claims, supporting evidence, and a table of mechanisms and reported effects. > [!NOTE] > > Consensus includes preprints by default, which keeps fast-moving fields current. Toggle them off in the filters if you only want peer-reviewed work. ## Step 2: Rank the approaches ```text theme={null} Which approach has the most evidence? Break down the methods and rank the key claims by strength of evidence. ``` Consensus re-reads the previous thread, weighs each approach, and ranks claims by how many papers support them and how thorough the studies are. ## Step 3: Compare specific papers Open **References**, tick three papers you want to go deeper on, and ask: ```text theme={null} Compare the different methods and results in each of these three papers. ``` With the papers attached, Consensus reads all three in full. Hover any citation to see the source passage. To see it in the PDF, click the paper's **Supporting Quotes**, open the **Attachment** tab, and pick the quote to highlight. ## Turn preprints off (or leave them on) Preprints keep fast-moving fields current but haven't been peer reviewed. They're labeled in results. To exclude them, open **Filter** and switch on **Exclude preprints** under General. The same panel has **Open access**, which limits results to papers with viewable full text. ![The General filter panel with the Exclude preprints and Open access toggles](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-03.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=de5170ab9adb3342b56e2faf79f8fa9d) ## Attach papers directly from the results Tick the boxes next to papers in the References list, then type your question in the search bar at the bottom. Only the attached papers are used, read in full where available. ![Search results with papers checked and attached as chips above the search bar](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/chat-with-full-text-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=a7f2f4f34725e380c8e449dccdc793b4) ## Follow-ups keep the thread's context Every follow-up in **Ask a follow up…** builds on the previous searches, so "which approach has the most evidence?" works without restating the topic. Follow-ups are available in Pro and Deep modes. To edit your first search instead, click **Filter** under the answer, adjust, and choose **Apply and resend**. ![The Ask a follow up box at the bottom of a Pro answer](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/threads-02.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=a5cd266a0bdbb2470c94d8db923604a5) ## Related tutorials > [card] [Filter for strong, recent studies](/search-basics/find-high-quality-recent-papers) > [card] [See how the field got here](/search-basics/map-field-history) # Filter for recent studies Source: https://docs.consensus.app/search-basics/find-high-quality-recent-papers Cut through preprints and low-impact work. Stack year, journal quality, and citation filters in plain language, then visualize the results on a research timeline. [How to Find High Quality Recent Research Papers](https://www.youtube.com/embed/wn0xzpsqbm8) When you search a topic, you usually get a flood of papers: preprints mixed with strong studies, low-impact work next to landmark trials. This tutorial shows how to get only credible, recent, high-impact research in two searches. **You'll learn how to:** 1. Filter results by year, journal quality, and peer-review status 2. Stack a citation threshold on top of an existing result set 3. Generate a research timeline to see when key papers were published ## Step 1: Search with filters in plain language Describe the topic and the filters you want in one message. Research Agent recognizes the criteria and applies the matching filters. ```text theme={null} Find high-quality papers on GLP-1 agonists and cardiovascular outcomes. Only papers since 2020, and only in Q1 journals. ``` Behind the scenes, Consensus: * Applies a **year ≥ 2020** filter and a **Q1 journal** filter * Expands your topic into the best search terms and runs two searches * Aggregates roughly 900,000 candidate papers, narrows them to 74 that meet your criteria, and shortlists the top 20 The AI overview on the left summarizes outcomes, subgroups, and comparators. The papers on the right are all post-2020 and all from high-quality journals. ## Step 2: Add a citation threshold and a timeline Ask a follow-up in the same thread. Research Agent keeps the context, so it re-reads the papers it just found rather than starting over. ```text theme={null} Narrow this down to papers with over 50 citations and show me a research timeline of when those papers were published. ``` Consensus reads the previous thread, keeps the papers with more than 50 citations (about 10), and builds a timeline showing when each was published and what milestone it represents. > [!TIP] > > Other filters you can state in plain language: study type (RCT, meta-analysis, systematic review), human studies only, minimum sample size, and specific countries. See [Research Agent](/core-features/research-agent). ## Every filter, explained Research Agent applies filters for you when you describe them, but you can also set them by hand. Click **Filter** in the search bar and a panel opens on the right. Pick your filters in any of its five sections (General, Methodology, Sources, Fields of study, and Countries), then close the panel. Filters stack, so "Q1 journals, RCTs only, human studies, past 5 years" all fit in one go. ![The Consensus search bar with the Filter button highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-02.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=37998cff44278c828c09de2e68ad365b) The Filter button sits at the right of the search bar. ![The Filter panel with General, Methodology, Sources, Fields of study, and Countries sections collapsed](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-01.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=e9b7d02e4a2ac115c0434b34a3c083b8) The Filter panel's five sections. Click one to expand it. > [accordion] General > > * **Publish year** — pick a preset (past 2, 5, or 10 years) or set a custom range. > * **Journal rank** — the SCImago Journal Rank (SJR) quartile. Q1 is the top 25% of journals in a field by influence; Q4 is the bottom 25%. > * **Citations** — a minimum citation count. High citations signal influence, not necessarily validity. > * **Exclude preprints** — remove papers that haven't been peer reviewed. > * **Open access** — only papers whose full text is viewable. > > ![The General filter panel with publish year, journal rank, citations, exclude preprints, and open access](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-03.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=de5170ab9adb3342b56e2faf79f8fa9d) > [accordion] Methodology: study design > > Consensus classifies every paper by study type. Filter by any of 19 designs across four groups: > > ![Study type categories: Experimental, Observational, Evidence Synthesis, and Other](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/study-types-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=f7a2f33710a5c9a0205f0f2d6de1a201) > > The four study-type groups and the designs in each. > > **Experimental** — the researchers test an intervention. > > * **RCT** — a randomized controlled trial with a randomly assigned control group. > * **Non-RCT experiment** — an intervention is given but participants aren't randomly assigned. > * **Bench experiment** — a laboratory study outside a clinical or field setting. > * **In vitro experiment** — conducted outside a living organism, such as a culture dish. > > **Observational** — subjects, events, or outcomes are examined without assigning an intervention. > > * **Field study** — data collected in a real-world setting. > * **Cross-sectional study** — a population at a single point in time. > * **Cohort study** — follows a group over time to study outcomes or exposures. > * **Longitudinal / panel study** — repeated observations of the same subjects. > * **Case-control study** — compares subjects with an outcome to those without, looking back for risk factors. > * **Case study** — an in-depth analysis of one person, group, event, or organization. > * **Case report** — describes a specific individual or small group, usually clinical. > * **Interview study** — qualitative data gathered through interviews. > * **Mixed methods study** — combines quantitative and qualitative methods. > > **Evidence synthesis** — reviews, summarizes, or statistically combines existing research. > > * **Systematic review** — a review with explicit, systematic inclusion criteria. > * **Meta-analysis** — a "study of studies" that statistically combines results. > * **Literature review** — a review without the explicit criteria of a systematic review. > > **Other** — simulation study, commentary, historical analysis. > > ![The Methodology filter panel listing all study designs](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-04.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=455791623111c76457e2bc8348ab4096) > > > [!TIP] > > > > A study type tells you how research was conducted, not how good it is. Systematic reviews, meta-analyses, and RCTs sit at the top of the evidence hierarchy for treatment questions, but the right design depends on the question. > [accordion] Methodology: study details > > * **Setting** — human or animal studies. > * **Sample size** — a minimum number of participants, subjects, or observations. > * **Duration** — how long the study ran, in days, weeks, months, or years. Useful for comparing short-term efficacy with long-term safety, dietary interventions over time, or short- versus long-term exposure. > * **Controlled studies only** — only studies with a control, placebo, or hold-out group. > > ![The study details filters: setting, sample size, duration, and controlled studies only](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-05.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=6e88d97e93d99f78a86e10c9548bda48) > [accordion] Sources, fields, and countries > > * **Sources** — filter by journal or by publisher (44 options), including Elsevier, Wiley, Springer Nature, Oxford University Press, IEEE, PLOS, BMJ, JAMA, and NEJM. > * **Field of study** — 23 fields, from Agricultural and Food Sciences to Sociology. > * **Countries** — the location of the study, from 236 countries. > > ![The Fields of study filter](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-06.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=e6c1b6e313cc642f7e5f9d9d501842db) > > ![The Countries filter](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/advanced-search-filters-07.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ac8dd34adf48b07e841c7db00be13e2a) Each of these is also available as a parameter in the [API](/api-get-started) and [MCP](/core-features/mcp). Full definitions live in the help center's [Advanced Search Filters](https://help.consensus.app/en/articles/9922799-advanced-search-filters) article. ## Related tutorials > [card] [Find the papers everyone cites](/search-basics/find-seminal-papers) > [card] [Catch up on a fast-moving field](/search-basics/catch-up-on-a-field) # Find popular papers Source: https://docs.consensus.app/search-basics/find-seminal-papers New to a field? Surface the foundational papers, deep-read the most cited one, and map its citation graph in three searches. [Finding the Seminal Papers in a Field](https://www.youtube.com/embed/bb6rox8IrZg) Entering a new field without a mentor, it's hard to know which papers everyone considers foundational. This tutorial uses the financial instability hypothesis in behavioral economics as the example. **You'll learn how to:** 1. Identify landmark papers in any discipline 2. Get a deep walkthrough of a foundational paper's core contribution 3. Visualize a citation graph to see a paper's place in the field ## Step 1: Ask for the seminal work ```text theme={null} Find the top papers on the financial instability hypothesis and explain why they're considered seminal. ``` Research Agent identifies the key author behind the theory, runs an author search for their work, and runs a broader search for other highly referenced papers. It reads the full text and abstracts of the strongest candidates and returns a table of key works, their core contributions, and why each counts as seminal. ## Step 2: Deep-read the most cited paper ```text theme={null} Pull up the most cited paper in this set and walk me through its core contributions. ``` Consensus uses the context from the first search, identifies the most cited paper (the original Minsky paper), reads its full text, and summarizes the core theoretical insights. Ask follow-ups for deeper analysis. ## Step 3: Open the citation graph Hover over the paper, open it, and click **Graph**. Consensus generates a citation graph of highly related papers that either cite this paper or are cited by it, both forward and backward in time. Hover any node to see its connections. > [!TIP] > > From here you can ask for a research timeline, a research gaps analysis, or similar papers from the last few years. See [Citation Graph](/core-features/citation-graph). ## Reading what Research Agent did Every Pro answer opens with a trace of the agent's work: how many papers it retrieved, how many were eligible, how many it included, and the sub-searches it ran with their filters. Expand it to check that the plan matched your intent before you trust the summary. ![A Research Agent trace showing retrieved, eligible, and included counts and the sub-searches run](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/research-agent-01.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=b6eef8e56acdc999c4b91cda8a2ca6e8) The agent's plan: two parallel searches with year and journal filters, 14.2K retrieved, 63 eligible, 20 included. ## Reading the citation graph ![A citation graph with three seed clusters connected by citation arrows](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/citation-graph-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=fa81f851d9a177afa1055c601ebfc3e4) Older papers sit left, newer right; node size is citation count; color marks each seed's cluster. | Element | Meaning | | - | - | | Horizontal axis | Time, older on the left | | Node size | Citation count | | Color | Seed clusters; gray nodes are bridge papers | | Opacity | How well connected the paper is | | Arrows | Point from a paper to the paper it cites | Use **+ Seed** to promote a result and regenerate, the minus button to exclude one, and **Density** to show 20, 40, or 60 papers. See [Citation Graph](/core-features/citation-graph) for how ranking works. ## Related tutorials > [card] [Follow a paper's citations](/search-basics/trace-citations) > [card] [See how the field got here](/search-basics/map-field-history) # Review a grant before you submit Source: https://docs.consensus.app/search-basics/grant-review Upload your grant and use a Deep Review to find similar studies, compare designs, identify literature gaps, and validate your effect size and sample size assumptions. [AI-Powered Grant Review: Validate Your Study Design Before You Submit](https://www.youtube.com/embed/zMIPTtJuvkM) Reviewers will challenge your sample size, your novelty claim, and your study design. Run that critique yourself first. The example uses a \~70-page grant on progesterone and muscle protein synthesis. **You'll learn how to:** 1. Upload a grant as an attachment 2. Use a Deep Review to find similar studies and compare designs 3. Validate your effect size and sample size against comparable trials ## Step 1: Upload the grant and ask for a comparison Click **Add attachment** in the search bar and upload the grant PDF. Toggle on **Deep**, then ask: ```text theme={null} Find studies with similar questions to this grant, compare the experiment design, study populations, and stated limitations, and find gaps in the literature today that actually address our proposal and justify why a grant might be provided here. ``` Deep Review reads the grant in chunks, identifies the core topic, and runs nearly 20 searches across the index. In the example it retrieved \~6 million candidates, found 150 relevant, and read the top 50. The answer covers the closest prior studies, a design comparison with their limitations, a critique of the population, and the gaps the proposal addresses. Hover any quoted citation to see the passage; a checkmark means it came from full text, not just the abstract. Click **Open** to jump to it in the paper. Open the sidebar to browse all 150 papers, switch to **table view**, or download them. ## Step 2: Critique the statistics ```text theme={null} Critique my effect size assumption and sample size. ``` Consensus compares the grant to similar approved studies and reports whether the assumptions hold, with a table of comparable trials showing design, effect size, and sample size. ## Privacy and sharing Uploads are private and aren't used to train models. To work with a lab, open **My Library**, put the grant and shortlisted papers in a Collection, and click **Share** to add collaborators. ## Uploading from the search bar Click **+** in the search bar, choose **Upload papers**, pick the file, then type your question. The upload lands in My Library too, so you can attach it again later. Text-based PDFs up to 50 MB; scanned PDFs can't be processed. ![The + menu in the search bar with the Upload papers button highlighted](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/upload-documents-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=3c8c53dbe89815f9cfb6d1f1c1e94ff8) ## Watch the plan as it runs The Deep Review panel shows every phase and search, with counts of papers identified, screened, eligible, and included. It's the fastest way to see whether Consensus understood the grant's core question. ![The Deep step panel showing retrieved, eligible, and included counts and the searches it ran](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/deep-search-steps.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=5339244798bc7c1366b0156fa02c9c58) ## Cite and export Click the thread settings button at the top of the page to choose a **Default export format** (APA, MLA, and others). Then the export button under the report gives **Copy text**, **Copy with citations**, or **Export as PDF**. ![The thread settings menu with the Default export format option set to APA](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/deep-review-03.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=f0ce442033e4a7ebc194d371362cf3ed) The default export format lives in thread settings. ## Share with your lab Open the Collection in My Library and click **Share**. Invite by email or copy an invite link; choose **Viewer** (can read and run searches on the Collection) or **Editor** (can also add and remove papers). Anyone you invite can see uploads inside the Collection. ![A Collection open in My Library, with the Share button in the top right highlighted](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/share-collection-01.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ab61c1adbfdbc0383aa1cc63d73ce496) > [card] [Consensus for grant writing](https://consensus.app/home/resources/grant-writing-consensus/) > > A full guide with sample prompts for each section of a proposal: background and significance, the gap, rationale, methods, outcome measures, and impact. ## Related tutorials > [card] [Check a manuscript's citations](/search-basics/manuscript-citation-check) > [card] [Grant finder](/use-cases/grant-finder) # Check a manuscript's citations Source: https://docs.consensus.app/search-basics/manuscript-citation-check Upload your manuscript and use a Deep Review to find supporting and challenging studies, surface papers you missed, and do a final gut check before submitting. [Never Miss a Citation Again: AI Manuscript Review with 400M Papers](https://www.youtube.com/embed/NxJ__xWXBT4) Nothing is more embarrassing than a reviewer pointing out a paper you should have cited. This tutorial uploads a manuscript (on TIMP1) and asks Consensus to find what supports it, what challenges it, and what's missing. **You'll learn how to:** 1. Upload a manuscript as an attachment 2. Run a Deep Review to find supporting, challenging, and must-cite papers 3. Verify each finding at the exact quote in the source paper ## Step 1: Upload and ask Click **Attachments › Upload papers** and select the manuscript. Toggle on **Deep** for the most thorough search, then ask: ```text theme={null} Can you look through this manuscript? What are relevant studies that support or challenge the findings? And what are papers that should definitely be cited alongside the study? ``` Deep Review reads the manuscript, extracts the core concepts, and runs multiple differently-phrased searches. It reads the initial results, decides what follow-up searches to run (in the example: the publishing lab's other work and the protein's role in the cell life cycle), and even crawls the reference list of a key paper. It found 413 eligible papers and read 50. ## Step 2: Read the answer The answer is organized as: * **Core takeaway** — what the manuscript reports and its main arguments * **Evidence strength** — claim by claim, with papers that support or refute each one * **Studies that support the manuscript** * **Studies that challenge or reframe it** * **Papers that should be cited** Click any claim to see the exact quotes from the source paper. Click **Open** to jump to that passage in the PDF. ## When to use it Early in writing, to find areas worth investigating. Late in writing, as a gut check before submission or before sending to a supervisor. > [!TIP] > > Uploaded papers stay in **My Library** so you can attach them again later. ## Two ways to upload * **From the search bar** — click **+**, then **Upload papers**, and ask your question in the same message. * **From My Library** — click **+ Add › Upload** to add up to 20 PDFs (or RIS/BibTeX files), then attach them to any search with **+**. ![The Add papers menu in My Library with Upload, Paste DOIs, and Import from Zotero](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/upload-documents-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=0599d0d6cf4cab7a24400a00b8b02d3e) Uploads must be text-based PDFs under 50 MB. Manuscripts, drafts, dissertations, protocols, and internal reports all work, and they stay private to you. ## Triage the 400 eligible papers The answer cites the 50 papers it read, but the References panel holds every eligible paper. Switch to **Table** view to compare them by population, methods, results, sample size, and more, or export the list as CSV or RIS. ![The References panel in table view](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/results-view-03.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=30c01abd7747743f621a731d6cecb58f) ## Export the review **Export** at the bottom gives you **Copy text**, **Copy with citations** in your reference style, or **Export as PDF**; tables and headings survive a paste into your writing tool. ![The Export menu at the bottom of a report](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/deep-review-04.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=ddb54a5a078f91f1dbff3598da7d5e1f) > [card] [Writing literature reviews with AI](https://consensus.app/home/resources/writing-literature-reviews-with-ai/) > > How to plan a review, find and manage sources in Consensus, read critically, and draft with confidence. ## Related tutorials > [card] [Review a grant before you submit](/search-basics/grant-review) > [card] [Ground a manuscript](/use-cases/ground-a-manuscript) # See how the field got here Source: https://docs.consensus.app/search-basics/map-field-history Get the historical arc of a field, not just a list of papers. Early foundations, key breakthroughs, and what changed in the last three years. [How to Map the History of a Research Field](https://www.youtube.com/embed/GmbNdGGwMpo) Entering a fast-moving field, you need to know what was foundational, what changed it, and where it stands now. This tutorial maps CRISPR gene editing from early work to clinical use. **You'll learn how to:** 1. Surface early foundational work in any field 2. Identify the breakthroughs that shifted the field 3. Generate a recent-only summary of what changed in the last 3 years 4. Read the full reference list in condensed or table view ## Step 1: Ask for foundations and breakthroughs ```text theme={null} Find the early foundational work on CRISPR as a gene editing tool and identify key breakthroughs that turned it into a clinical tool. ``` Research Agent runs searches across different time periods, then notices it needs landmark papers to fill gaps. It identifies papers referenced by the key results, looks them up directly, reads 20 of them, and returns a timeline of milestones and the papers to read. Hover a citation to open the paper. Open-access papers load directly inside Consensus. Click **References** in the sidebar to see the full list as a condensed view or a table view that includes the methods used. ## Step 2: Ask what changed recently ```text theme={null} What's changed in the last 3 years and what findings really stand out as the strongest? ``` Consensus re-runs the searches with a 3-year recency filter and summarizes the key claims and evidence, ranking strength by citation count, study type, and how many papers carry the finding forward. In the example: clinical breakthroughs in sickle cell disease and the emergence of next-generation gene editing. ## Browse the full reference list The answer cites the papers that best support each claim, but the thread holds more. Click **References** in the top right and use the dropdown to see **All cited papers** or the papers **By message**. Switch between **Default** (with key takeaways), **Compact**, and **Table** views with the buttons above the list. ![The References panel showing papers cited across a thread](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/threads-04.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=8958b4f28aedf48631e12089836fe5e5) ![Search results in default view with key takeaways and study-type badges](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/results-view-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=feb9f3161478d73964292681c03016d8) Default view shows a key takeaway under each paper; Compact hides it so more papers fit on screen. ## Export the reading list Click the download arrow at the top of the results to export **CSV**, **RIS** for your reference manager, or a formatted **Bibliography** (APA, MLA, Chicago, Harvard, BibTeX, AMA/Numeric). ![The export menu on the results panel](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/pro-messages-05.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=34fefbf55ef6088db4d48d2e849557c6) ## Related tutorials > [card] [Find the papers everyone cites](/search-basics/find-seminal-papers) > [card] [Catch up on a fast-moving field](/search-basics/catch-up-on-a-field) # Add your own PDFs Source: https://docs.consensus.app/search-basics/private-documents Upload internal reports, proprietary data, or unpublished documents into a Collection and search them alongside 400M+ scholarly articles, privately. [Search Your Private Documents Alongside 400M Research Papers](https://www.youtube.com/embed/eMBzOElq4Y8) Internal reports and unpublished documents can live next to the published literature. This tutorial creates a Collection of market reports and asks a question that draws on both. **You'll learn how to:** 1. Create a Collection and upload private documents 2. Run a search that combines your documents with published research 3. Share the Collection with a team ## Step 1: Create a Collection and upload Open **My Library**, create a new Collection (the example is called *Economic reports*), and upload your files. Peer-reviewed papers and internal documents both work. Consensus indexes each document so it understands what it's about. Uploads are private, not shared with anyone, and not used to train models. ## Step 2: Ask a question that spans both Start a new search: ```text theme={null} What does the research say on the trends in the battery market in China in 2026? ``` Consensus checks the Collection and finds the relevant uploaded report, then runs broader searches across the published literature. In the example it read two full texts and analyzed 20 abstracts, and the overview includes the uploaded report alongside the published papers. You can also scope a question to only the Collection by attaching it to the search bar. ## Step 3: Share with your team Open the Collection and click **Share**. Invite people by email as a **Viewer** (can use the Collection in searches) or an **Editor** (can also add documents). > [!NOTE] > > For larger teams, see the Team plan on the [pricing page](https://consensus.app/pricing/), or contact Consensus for institutional setup. ## What you can upload Published papers, drafts, manuscripts, experiment proposals, dissertations, clinical protocols, private research reports, and other research documents. Up to 20 PDFs at a time, 50 MB each, text-based. RIS and BibTeX files import their references. ![The Add papers menu in My Library](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/upload-documents-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=0599d0d6cf4cab7a24400a00b8b02d3e) ## Pick exactly which documents to use Click **+** in the search bar to open **Add sources** and tick specific documents or papers from your Library, or attach a whole Collection. Leave nothing attached and choose **Corpus › All** to search the full index alongside your Library, as in the example above. ![The Add sources panel with several documents checked](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/upload-documents-03.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=d92235ab84d81c718e5b0342f8ac7bc2) The Add sources picker: choose individual documents or upload new ones. ## Sharing, in detail | Role | Can | | - | - | | Viewer | View papers and sub-Collections; run searches on the Collection | | Editor | Everything a Viewer can, plus rename, add sub-Collections, add and remove papers | * Invite by email or copy an invite link with a preset role. * People without an account are prompted to create a free one. * Shared Collections show a person icon and an **Added by** column; recipients find them under **Shared with me**. * Editors can remove papers from the shared Collection, but those papers stay in your own Library. * Saved threads inside a Collection aren't shared yet. ![The Shared with me entry in the My Library sidebar](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/share-collection-02.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=a3bcec599adc4d4cd7b7b26fd7e5aa6d) ## Related tutorials > [card] [Library](/core-features/library) > [card] [Find research gaps](/learn/find-research-gaps) # Look up a researcher's work Source: https://docs.consensus.app/search-basics/research-an-author Map a researcher's publications, find their most influential paper, and trace its downstream impact. Useful before a meeting, a hire, or a citation. [How to Research an Author's Body of Work and Most Influential Papers](https://www.youtube.com/embed/7aoHQoOv0JY) Sometimes you need to understand one person's footprint: what they've published, and what influence it's had. Instead of scouring ResearchGate or Google Scholar, ask Consensus. The example uses a PhD candidate with a handful of papers, not a famous name, to show it works for early-career researchers too. **You'll learn how to:** 1. Run an author-scoped search filtered by topic 2. Rank a researcher's papers by citation impact 3. Visualize the citation graph and regenerate it with several seeds ## Step 1: Find the author's papers on a topic ```text theme={null} Find papers by Alysha D'Souza on muscle development in female athletes and identify the most influential paper in that set. ``` Research Agent resolves the correct author, finds their papers, narrows a broad set of \~12,000 results to the ones where they're a primary author or co-author on the topic, and identifies the most influential one. In the example, that's an umbrella review with 102 citations. Click through to the DOI to confirm authorship. ## Step 2: Trace the paper's impact ```text theme={null} Show me later work that was built on that paper and summarize its impact. ``` Consensus walks the citation graph forward, finds \~55 eligible citing papers, shortlists 20, and summarizes how the field responded. In the example, the review exposed methodological flaws, and later studies moved from calendar-based to hormone-informed cycle tracking. ## Step 3: Triangulate with the citation graph Scroll up to the original thread, click the paper, and click **Graph**. A new tab opens with the 15 papers most related to it, forward and backward. Open the sidebar, add two or three more papers as seeds, and click **Regenerate graph**. The graph now shows papers related to all of the seeds together, which is a fast way to find work on a narrow niche. ## Reading and refining the graph ![A citation graph with three colored seed clusters](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/citation-graph-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=fa81f851d9a177afa1055c601ebfc3e4) Each seed gets a color; papers one hop away take that color. Gray nodes bridge clusters. * **Horizontal axis** is time; **node size** is citation count; **opacity** shows how connected a paper is. * **Arrows** point from a paper to the paper it cites. Hover a node to highlight its connections. * **+ Seed** promotes a result to a seed; the minus button excludes it; **Density** switches between 20, 40, and 60 papers. * Click **Chat** at the bottom to ask questions with all seeds and results attached. Graphs aren't saved yet (beta), so promote seeds and save the papers you care about to your Library before closing the tab. ## Related tutorials > [card] [Follow a paper's citations](/search-basics/trace-citations) > [card] [Citation Graph](/core-features/citation-graph) # Grow a reading list from a paper Source: https://docs.consensus.app/search-basics/snowball-reading-list Start from one paper you trust and triangulate similarity, citations, and references to find every adjacent study, then map the gaps. [How to Snowball From One Research Paper to a Full Reading List](https://www.youtube.com/embed/YvrcZbLCips) One trusted paper is enough to find the literature around it. Three signals help: papers that look similar, papers that cite it, and papers it cites. This tutorial uses Cryan & Dinan (2012), *Mind-altering microorganisms*, on the gut microbiome and depression. **You'll learn how to:** 1. Find semantically similar papers from a single seed 2. Build a research timeline from the collected set 3. Generate a research gaps matrix showing what topics and populations are missing ## Step 1: Find the seed and its neighbors ```text theme={null} Find the paper "Mind-altering microorganisms" (Cryan & Dinan, 2012), then find other similar papers. ``` Research Agent locates the paper, searches on signals from its abstract and full text for semantically similar work, and walks the citation graph in both directions. From over a million candidates it deduplicates to 99 close matches and analyzes the top 20. Scroll the sidebar and click **Show more** for the full list. ## Step 2: Visualize coverage and gaps ```text theme={null} Create a research timeline and a research gaps matrix showing what the collected studies cover or miss. ``` Consensus builds a timeline of when each paper was published relative to the seed, then a gaps matrix. In the example, only one paper covers human clinical trials of short-chain fatty acids and metabolites. ## Step 3: Fill the gaps ```text theme={null} Find me papers from the last 5 years that fill in the gaps mentioned. ``` Consensus runs a parallel search for each major gap, with a 5-year recency filter on every one. > [!TIP] > > Hover any paper and click the **bookmark** to save it to your [Library](/core-features/library) as you go. ## Save papers as you find them Click the **bookmark** on any paper to save it, or the bookmark at the top of the page to save the whole thread. Pick an existing Collection or create a new one. Collections can be nested by choosing a parent under **Create in**. ![A search result with the save-paper bookmark highlighted and the Collection picker open](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/my-library-04.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=eb05f747739795b7c575d33c3f4d253b) The bookmark on a paper and at the top of the thread. ## Review every paper the thread used Click **References** in the top right. The dropdown switches between **All cited papers** (everything across the thread) and **By message** (papers behind one specific search). Tick papers here to save several at once or attach them to a follow-up. ![The References panel with the All cited papers / By message dropdown open](https://mintcdn.com/consensus-f7e6ffce/k_-rK_7yzAtjVz-H/images/help/threads-04.png?fit=max\&auto=format\&n=k_-rK_7yzAtjVz-H\&q=85\&s=8958b4f28aedf48631e12089836fe5e5) ## Related tutorials > [card] [Follow a paper's citations](/search-basics/trace-citations) > [card] [Find research gaps](/learn/find-research-gaps) # Follow a paper's citations Source: https://docs.consensus.app/search-basics/trace-citations Understand where a paper came from and how the field built on it. Bi-directional citation tracing using Attention Is All You Need as the anchor. [How to Trace Citations Forward and Backward From Any Research Paper](https://www.youtube.com/embed/KVCVOqndX2U) When you find an important paper, you want to know what built up to it and how the field built on top of it. This tutorial uses *Attention Is All You Need* (Vaswani et al., 2017) as the anchor. **You'll learn how to:** 1. Pull a paper's outgoing references grouped by intellectual lineage 2. Trace forward citations and filter them by topic and recency 3. Map the past and future of any foundational paper in one workflow ## Step 1: Look backward ```text theme={null} Show me the papers that "Attention Is All You Need" cites. What intellectual lineage did the authors build upon? ``` Research Agent looks up the paper, crawls its references through the citation graph, shortlists and reads the most relevant ones, and returns the lineage: key attention precursors, the architectures that evolved over time, and a summary of the field. ## Step 2: Look forward, narrowed to a niche ```text theme={null} Show me papers from the last 5 years that cited it for speech and audio modeling. How did they adapt attention to spectrograms or raw audio, and what tasks benefited? ``` Consensus runs several searches on how the field developed, applies the recency filter, explores the citation graph for papers that reference the anchor, and runs extra searches to fill gaps. The result is an overview of how one seminal paper shaped a specific subfield in a specific window. > [!TIP] > > For a visual version of the same idea, open any paper and click **Graph**. See [Citation Graph](/core-features/citation-graph). ## The same trace, as a graph Citation Graph does this visually. It ranks papers by three citation signals rather than keywords: * **Direct citation** — the paper cites your seed, or your seed cites it. * **Co-citation** — how often the paper and your seed appear together in other papers' reference lists. * **Bibliographic coupling** — how many references the paper shares with your seed. A traditional crawl only looks forward or backward from one paper; combining the three signals surfaces related work a crawl misses. ![A citation graph showing seed papers and their citing and cited neighbors over time](https://mintcdn.com/consensus-f7e6ffce/wHajK_wM4l0mxqJf/images/help/citation-graph-01.png?fit=max\&auto=format\&n=wHajK_wM4l0mxqJf\&q=85\&s=fa81f851d9a177afa1055c601ebfc3e4) Arrows point from a paper to the paper it cites. Hover a node to highlight its connections. Open it from any paper's sidebar with the **Citation Graph** button, or from **Graph** in the left nav. Up to 5 seeds; 20, 40, or 60 papers. ## Related tutorials > [card] [Grow a reading list from one paper](/search-basics/snowball-reading-list) > [card] [Look up a researcher's work](/search-basics/research-an-author) # What you can build with the Consensus API and MCP Source: https://docs.consensus.app/use-cases Search 400+ million research papers and read the full text of open-access ones, for internal tools, scientific agents, and custom research systems. Consensus gives your code and your AI tools search across **400+ million scholarly articles, including 350 million peer-reviewed research papers**, matched on full text where Consensus holds it, not just titles and abstracts. Ask in natural language, narrow with the filters a researcher would apply (study design, date, journal quality, field, country), and get back ranked papers, the links you need to cite them, and, for open-access papers, the passages that answer your query or the complete text. ## Full text, not just abstracts Methods, doses, and effect sizes rarely make it into the abstract. Consensus searches the whole paper and, for open-access papers, returns the part that matters. * **Search inside the paper:** match on methods and results, not just abstracts. * **Return the passage:** get query-relevant excerpts from open-access papers, labeled by section. * **Read the whole paper:** get the [complete text](/api-full-text#complete-text) of an open-access paper as Markdown. * **Ground every answer:** cite the actual text, not a paraphrase. > [accordion] See a real full-text response > > On all paid plans, searches return full-text excerpts from open-access papers. Nothing extra goes in the request: > > ```bash theme={null} > curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ > --data-urlencode "query=what dose of creatine improves working memory" > ``` > > ```json wrap theme={null} > { > "results": [ > { > "title": "Creatine and improvement in cognitive function: Evaluation of a health claim pursuant to article 13(5) of regulation (EC) No 1924/2006", > "authors": ["Dominique Turck", "T. Bohn", "Montaña Cámara", "…"], > "journal_name": "EFSA Journal", > "publisher_name": "Wiley", > "publish_year": 2024, > "sjr_best_quartile": 1, > "citation_count": 5, > "doi": "10.2903/j.efsa.2024.9100", > "url": "https://consensus.app/papers/creatine-and-improvement-in-cognitive-function-turck-bohn/7e703e828b815257b31687a25eceac9d/", > "takeaway": "Creatine supplementation has not been established as a cause-and-effect relationship for improving cognitive function in one or more domains.", > "full_text_chunks": [ > "Section: Human intervention studies in healthy individuals conducted under normal conditions | Two studies had an initial loading phase of 20 g/day of creatine for 5 days, followed by 5 g/day … A significant improvement in working memory, as measured by the Backwards Digit Span test, was observed only in the study of shorter duration …", > "Section: Weighing of the evidence | … the acute effect of creatine on working memory reported in some studies at doses of 20 g/day given for 5-7 days … was not observed at lower creatine doses, ranging from 2.2 to 14 g/day …" > ] > } > ] > } > ``` > > A real response, trimmed to one result with shortened excerpts. The `takeaway` gives you the paper's conclusion. The `full_text_chunks` give you what the abstract leaves out — the doses, durations, and study designs behind that conclusion — each labeled with the section of the paper it came from. ## What people build Most teams building on Consensus fall into one of two groups. > [card] [Internal tools for your organization](/use-cases/internal-tools) > > Enterprise R\&D teams, biotechs, and AI startups connecting paper search to internal assistants, scientific agents, monitoring agents, and literature tables. > [card] [Custom research systems](/use-cases/research-systems) > > Academics and research groups building their own literature-review, writing, and reporting pipelines — in code or inside the AI tools they already use. ### Internal tools for your organization [Paper search for your agent Ground the scientific agent you ship in full-text papers, the way Phylo and Owkin do. See how ](/use-cases/internal-tools#give-your-scientific-agent-paper-search) [Field-monitoring agent Surface what's new in your team's fields every week, and track who's publishing it. See how ](/use-cases/internal-tools#keep-on-top-of-a-field) [Self-filling literature table Run the same literature questions across hundreds of targets or compounds, on demand or on a sync. See how ](/use-cases/internal-tools#build-literature-tables-that-fill-themselves) More in the [internal tools guide](/use-cases/internal-tools), including internal AI assistants and exhaustive paper lists. ### Custom research systems [Literature review pipeline Fan out, filter, and deduplicate into a review you can reproduce and report. See how ](/use-cases/research-systems#literature-review-pipelines) [Citation-grounded writing Back every claim in a draft with a real paper, or flag it as unsupported. See how ](/use-cases/research-systems#writing-with-grounded-citations) [Consensus in your AI assistant Run the same workflows in Claude or ChatGPT over MCP, with no code. See how ](/use-cases/research-systems#use-the-tools-you-already-have) More in the [research systems guide](/use-cases/research-systems), including scheduled evidence reports. ### Real examples > [card] [Phylo: literature retrieval for agentic biology](https://consensus.app/home/community-voices/phylo/) > > Phylo's agents fan each biology question out into parallel Consensus API searches. One internal team screens about 20,000 targets across hundreds of diseases, with almost zero production errors. > [card] [Owkin: a full-text evidence layer for K Pro](https://consensus.app/home/community-voices/owkin/) > > Owkin replaced a hand-maintained set of PubMed abstracts with Consensus MCP inside its agentic AI scientist — 25x the paper coverage, shipped in days. ## Two ways to connect You can reach Consensus two ways: > [card] [REST API](/api-get-started) > > One `x-api-key`, every filter, and pagination to exhaustion. For pipelines, internal tools, and anything scheduled. > [card] [MCP server](/consensus-mcp) > > Plug into Claude, ChatGPT, Claude Code, or any MCP-compatible agent. No code, and each user signs in with their own account. ## Explore more > [card] [Get started](/use-cases/get-started) > > Pick a surface, run your first filtered search, and find the guide for what you're building. > [card] [Use case library](/use-cases/gallery) > > 30 worked examples you can filter by persona, each with the full prompt or API call behind it. > [card] [Best practices](/use-cases/best-practices) > > The five building blocks, rate limits, and the grounding rules that keep output trustworthy. # Screen abstracts against inclusion criteria Source: https://docs.consensus.app/use-cases/abstract-screening Apply explicit inclusion and exclusion criteria across thousands of abstracts, with a decision log a second reviewer can audit. ## What you get One row per record: include, exclude, or flag for human review — each with the criterion that decided it and a supporting quote. The most labour-intensive step in evidence synthesis, turned into something a human checks rather than performs. > [!NOTE] > > Consensus gives you the primitives for this — filters, pagination, abstracts, and study types. It is not a packaged screening product, so you are building the workflow yourself. Customers ask us for this regularly; treat the recipe below as the current best path. ## Who it's for Anyone running a systematic or scoping review — graduate students, methodologists, evidence synthesis units. Traditionally two people screen every abstract independently, which is where the weeks go. ## The prompt ```text theme={null} You are a systematic review screener. Apply the criteria below to each record. Do not summarize the papers. Do not decide what would be interesting. Inclusion criteria: [ONE PER LINE] Exclusion criteria: [ONE PER LINE] For each record, work through the criteria in order and stop at the first one that decides the case. Then output exactly one verdict: - INCLUDE — meets every inclusion criterion, triggers no exclusion. - EXCLUDE — fails a specific criterion. Name which one. - FLAG — the abstract does not contain enough information to decide. Say which criterion you could not evaluate and what you would need to read. Rules: - FLAG is not a failure state. A record whose abstract omits the sample size cannot be excluded for sample size — flag it. - Never infer a study design the abstract does not state. If it does not say randomized, it is not an RCT for screening purposes. - Quote the phrase that decided the verdict. A verdict with no quote is not auditable and will not survive a second reviewer. - Judge the record in front of you. Do not use anything you know about the paper from elsewhere. Output: Record ID | DOI | Verdict | Deciding criterion | Supporting quote | Confidence (high/low). ``` ## Calibrate before you scale **1. Have a human screen 100 records first** Screen a random sample by hand, then run the agent over the same 100 and compare. You are measuring agreement, and more importantly reading the disagreements. **2. Fix the rubric, not the model** Almost every disagreement traces to a criterion two humans would also read differently. Rewrite the criterion until it is unambiguous, then re-run. **3. Watch the FLAG rate** A near-zero flag rate means the agent is deciding cases it cannot actually decide from an abstract. Somewhere between 5% and 20% is normal depending on how much your criteria depend on methods detail. **4. Escalate borderline records to full text** For flagged records, search again with the criterion as the query, and screen the excerpts in `full_text_chunks` instead. Paid plans return them for open-access papers, chosen against your query. ## Narrowing before you screen Every filter you apply is a record you do not have to screen — but each one also risks dropping an eligible study, so record what each rung costs you. ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=resistance training older adults muscle strength" \ --data-urlencode "study_types=rct" \ --data-urlencode "human=true" \ --data-urlencode "sample_size_min=30" \ --data-urlencode "page_size=200" ``` Use the [filter ladder](/use-cases/best-practices#filter-ladder) and keep the count at each rung, because a reviewer will ask. ## Audit requirements Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability). Screening specifically needs: total records retrieved, records after deduplication, screened, included, excluded by criterion, and flagged. Excluded counts must be reported per criterion — PRISMA requires reasons, not a single total. ## What to check before you trust it * **Never let the agent be the only screener on a published review.** It is a first pass that a human adjudicates, and journals will ask who screened. * **Check the excluded pile, not just the included one.** False exclusions are invisible in the output and fatal to the review. * **Re-screen after any criterion change.** A mid-run rubric edit makes the earlier verdicts incomparable. ## Related > [card] [Build a PRISMA-ready search corpus](/use-cases/prisma-corpus) > > The step before screening: assembling the records. > [card] [Write the search-strategy section of your methods](/use-cases/search-strategy-methods) > > Turn the log into the paragraph reviewers require. > [card] [Best practices](/use-cases/best-practices) > > Filter ladder, grounding rules, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Add a citable evidence layer to an agent you already ship Source: https://docs.consensus.app/use-cases/add-an-evidence-layer Give a working agent access to peer-reviewed literature as a tool, so its claims carry citations instead of assertions. ## What you get An agent that already does something valuable, now grounded: every claim it makes traces to a real paper, and "the literature does not answer this" becomes an answer it can give. The integration is a tool registration, not a re-architecture. Owkin shipped theirs to users in days. ## Who it's for Product engineers whose agent works but overclaims — it reasons well and cites nothing, and your users have started noticing. ## How it went at Owkin [Owkin](https://consensus.app/home/community-voices/owkin/) exposed Consensus to K Pro's orchestrator the same way it exposes its other tools. That architectural fit is the whole story: no new retrieval service, no corpus to maintain, no bespoke ranking layer. > "The Consensus MCP slotted directly into our agentic stack and gave K Pro's capabilities a dramatic upgrade overnight." > > — **Lucas Brechot**, Senior Product Manager, AI Products, Owkin Agents now ground their reasoning in full-text literature with programmatic controls for recency, citation count, and journal reputation — the filters a human researcher applies instinctively when judging whether a paper is relevant and credible. ## Wiring it in ```bash MCP theme={null} # OAuth per user, so each person's own plan and limits apply https://mcp.consensus.app/mcp ``` ```bash REST theme={null} # one service key, full control over pagination curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=your agent's search string" ``` Choose by auth model rather than by preference: **MCP** when each user should search under their own account and limits, **REST** when the integration needs one shared quota and programmatic pagination. ## What separates a grounded agent from a decorated one **1. Decompose before retrieving** One search per user question is the most common cause of thin answers. Split the question into subquestions and give each its own search with its own filters. **2. Retrieve excerpts, not just abstracts** On paid plans, MCP and the API both return full-text excerpts from open-access papers. Pass them to the generation step so it reasons over results paragraphs rather than the introduction. **3. Enforce grounding at generation time** Require a `doi` or `url` on every claim and drop unsupported sentences. Make "the literature does not answer this" a first-class output rather than a failure state. **4. Surface the filters you applied** Show the study types, years, and quality thresholds used. A researcher cannot trust a result set whose inclusion criteria are hidden. **5. Cache on the full parameter set** Cache on every parameter, not the query string alone, with a short TTL so new literature still lands. ## What to check before you trust it * **Check the unanswerable case.** Ask your agent something the literature genuinely does not settle. If it produces a confident cited answer anyway, your grounding rule is decorative. * **Spot-check citations against claims.** A real paper attached to a claim it never made is the failure that survives review. * **Watch your rate limit under real traffic.** Per-user OAuth spreads load across accounts; a shared service key concentrates it. See [rate limits](/use-cases/best-practices#rate-limits-and-result-caps). * **Do not hide the filters.** If your product narrows to Q1 journals by default, say so in the UI. ## Related > [card] [Swap a hand-maintained paper corpus for full-text search](/use-cases/swap-a-paper-corpus) > > When you already have a corpus and want it gone. > [card] [Answer one question with many parallel searches](/use-cases/parallel-search-angles) > > The retrieval pattern that makes a grounded agent good rather than merely cited. > [card] [Best practices](/use-cases/best-practices) > > Rate limits, grounding rules, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Benchmark literature providers against your own ground truth Source: https://docs.consensus.app/use-cases/benchmark-retrieval Build a curated query set with known-relevant papers and measure which retrieval provider actually surfaces them. ## What you get A defensible answer to "which literature API should we build on", measured on your own domain rather than on a vendor's marketing page — and a regression harness you keep afterwards. ## Who it's for The engineer choosing a retrieval provider, and whoever has to justify that choice to a technical buyer or a research team who will notice if it is wrong. ## How Phylo did it Before putting anything into production, [Phylo](https://consensus.app/home/community-voices/phylo/) ran a head-to-head benchmark. The team curated biomedical queries spanning several disease areas, each with a ground-truth list of papers already known to be highly relevant, then measured **at top 20 results** how many of the truly relevant papers each system actually surfaced. > "We benchmarked across many different providers. We ended up choosing Consensus." > > — **Kexin Huang**, Co-founder & CEO, Phylo **1. Curate queries across your real domain** Questions should span the disease areas, research contexts, or subfields your product actually serves — including the awkward ones. A benchmark built only from easy queries measures nothing. **2. Build ground truth per query** For each query, list the papers a domain expert would insist appear. This is the expensive part and the part that makes the result trustworthy; do not shortcut it with model-generated relevance labels. **3. Measure recall at a fixed k** How many ground-truth papers appear in the top k. Phylo used top 20. Fix k before you run anything, because choosing it afterwards is how benchmarks get massaged. **4. Hold the query text constant across providers** Different query phrasings favour different engines. Same strings, same k, and note where a provider needs different phrasing to perform — that is a real integration cost. ## Running it against Consensus ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=" \ --data-urlencode "page_size=20" \ --data-urlencode "include_semantic_score=true" ``` `include_semantic_score=true` returns a relevance score for the top 50 results, which lets you check score calibration alongside raw recall. It is API-only — the MCP `search` tool does not expose it. > [!NOTE] > > Benchmark unfiltered first. Filters change recall substantially, and a comparison where one provider is filtered and another is not measures your configuration rather than the engines. Add filters in a second pass to see how much precision they buy. ## Keep it as a regression harness The benchmark is more valuable after the decision than during it. Re-run it when you change prompts, filters, or query construction — it is the only way to tell whether a retrieval change actually helped or just felt better. Split the score in two: **retrieval quality** (did the right papers come back) and **grounding faithfulness** (did the answer stay true to them). A pipeline can fail at either, and the fixes are unrelated. ## What to check before you trust it * **Beware ground truth built from one provider's output.** If your labels came from a search on system A, system A wins by construction. Build them from expert knowledge or published reviews. * **Report recall per query, not just the mean.** One catastrophic query hidden inside a good average is exactly the failure your users will hit. * **Re-check page size.** Comparing top-20 against a provider capped below 20 is not a fair test — confirm the `page_size` echoed back matches what you asked for. * **Say when the difference is noise.** With 30 queries, small gaps are not signal. ## Related > [card] [Swap a hand-maintained paper corpus for full-text search](/use-cases/swap-a-paper-corpus) > > What to do once the benchmark decides it. > [card] [Screen thousands of target-disease pairs for evidence](/use-cases/high-throughput-screening) > > The workload that makes retrieval quality worth measuring. > [card] [Best practices](/use-cases/best-practices) > > Rate limits, grounding rules, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Best practices Source: https://docs.consensus.app/use-cases/best-practices The five primitives every Consensus workflow is built from, and how to tell a good use case from a poor one before you build it. Every use case in the [library](/use-cases/gallery) — and every build in the [internal tools](/use-cases/internal-tools) and [research systems](/use-cases/research-systems) guides — is a combination of five primitives. If you build only one thing, build these — then compose them. ## The five patterns | Pattern | What it does | How | | - | - | - | | **Fan-out and merge** | Turns one question into 5–30 searches and one clean corpus | Issue queries one per second, deduplicate on `doi`, keep which query found what | | **Filter ladder** | Trades recall for precision in controlled steps | Start unfiltered, then add `study_types`, then `human` and `controlled`, then `sjr_max` — record counts at each rung | | **Exhaustive sweep** | Gets every record, not the first page | Walk `page` from `0` while `is_end` is false, following `next_page`, at your plan's max `page_size` | | **Excerpt-first extraction** | Extracts numbers you can defend | Read `full_text_chunks` (paid plans) and phrase `query` as the thing being extracted, then keep each value with its excerpt | | **Diff against last run** | Turns any search into a monitor | Window with `year_min` and `month_min`, diff on `doi` against stored state, report only the delta | > [!TIP] > > The filter ladder is the highest-leverage habit. Most disappointing Consensus results are an unfiltered query, and most empty ones are four filters applied at once. Move one rung at a time and keep the counts. ## Fan-out and merge The single biggest quality difference between a thin answer and a good one. One query is one search; a real question deserves 5–30. **1. Decompose before retrieving** Split the question into subquestions — mechanism, population, outcome, competing explanations. Generate several phrasings per subquestion covering synonyms and adjacent terminology. **2. Pace them to your rate limit** Every query is independent, but there is a one-query-per-second rate limit on every plan except Enterprise. Unbounded parallelism gets you 429s, not speed — run searches sequentially and build the pacing into your system. See [rate limits](#rate-limits-and-result-caps) below. **3. Deduplicate on DOI** Merge on `doi`, keeping the first-seen record plus the list of queries that found it. How many queries surfaced a paper is a useful relevance signal in itself. **4. Keep the pre- and post-merge counts** You will need both to describe your coverage, and a sharp drop tells you your phrasings were near-duplicates. ## Filter ladder ```bash theme={null} # rung 0 — gauge volume curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=resistance training older adults muscle strength" # rung 1 — design --data-urlencode "study_types=rct,meta-analysis" # rung 2 — population --data-urlencode "human=true" --data-urlencode "sample_size_min=50" # rung 3 — quality --data-urlencode "sjr_max=1" --data-urlencode "exclude_preprints=true" ``` Record the result count at each rung. When a rung drops you to near-zero, that is the constraint to relax — and knowing which one it was is most of the diagnosis. ## Exhaustive sweep ```python theme={null} def sweep(query, **filters): page, out = 0, [] while True: body = search(query=query, page=page, **filters) out.extend(body["results"]) if body["is_end"] or body.get("next_page") is None: break page = body["next_page"] if (page + 1) * body["page_size"] > 1000: # a query returns at most 1,000 results break return out ``` > [!NOTE] > > Pagination past the first page requires a paid plan. `page_size` defaults to 20 and caps at 20 on Free, 300 on Pro and Teams, 750 on Deep, and 1,000 on Enterprise; a request over your ceiling is clamped rather than rejected, so check the `page_size` echoed back in the response. On every plan a single query returns at most 1,000 results in total, because `(page + 1) × page_size` can't exceed 1,000 — reach wider coverage with more queries, not more pages. ## Rate limits and result caps Two separate ceilings shape every workflow here, and confusing them is the most common cause of a pipeline that quietly under-reports. **Rate limit — how fast you may ask.** The API allows one search per second on every plan except Enterprise. MCP allows three per second on paid plans: | Plan | API requests per second | MCP requests per second | Max papers per request (API / MCP) | | - | - | - | - | | No account | — (the API needs a key) | 5, shared across all users | — / 3 | | Free | 1 per user | 1 per user | 20 / 10 | | Pro | 1 per user | 3 per user | 300 | | Deep | 1 per user | 3 per user | 750 | | Teams | 1 per user | 3 per user | 300 | | Enterprise | Custom | Custom | 1,000 | Full details are in [API plans and access](/api-plans-and-access) and [MCP plans and access](/mcp-plans-and-access). > [!WARNING] > > Do not fan out with unbounded parallelism. Run searches one at a time and pace them, on every plan. Without an account the limit is shared across all users, so it is the least predictable option — sign in before building anything real. Retrieving details for a paper you have already found is not the bottleneck — searching is. **Result cap — how much comes back.** A response that says it found more papers than it returned has been capped by your plan. That distinction matters: **you can only cite what was shown**, and a thin result set may reflect your plan ceiling rather than a genuine gap in the literature. Parse both numbers, and when they differ, say so in your output rather than reporting the shown count as the whole literature. **Handling 429s.** The two kinds mean different things: * **Per-second rate limit** — there is a one-question-per-second limit on every plan except Enterprise. Account for it in whatever you are building rather than firing many searches in parallel: pace your calls, honour `retry-after`, and add jitter. Retrying helps. * **Monthly search allowance** — you have used the searches included in your plan. Retrying does not help. Turn on additional usage in your account settings to keep searching past the included allowance. A workable retry policy, and the one the Consensus-built skills use: on failure wait 3 seconds and retry once; after 3 consecutive failures across any combination of calls, stop and surface the problem rather than silently producing a thin result. ## Grounding and auditability Every workflow in the gallery ends in something a person will act on — a citation in a manuscript, a target decision, a claim that ships. These five rules are what keep that output trustworthy, and they are the ones the Consensus-built skills enforce most strictly. **1. Cite only what came back in this session** Never supplement results with model knowledge, a previous conversation, or a paper you happen to know. A fabricated citation in an academic or regulatory context costs far more than a thin result set. If you include something not retrieved in this session for context, label it explicitly and exclude it from every count. **2. Count three things separately** **Queries sent**, **results returned**, and **results cited**. These are different numbers and conflating them is how a five-paper review gets described as comprehensive. Track all three from the first search, not retroactively. **3. Detect the plan cap and report it** When a response reports finding more papers than it returned, your plan capped it. Only the returned papers are citable. Surface it: a sparse section may reflect your ceiling rather than a genuine gap in the literature, and the reader cannot tell the difference unless you say which it was. **4. Confirm each result before moving on** A step is not done because you sent the request — it is done when the response is back and contains data. On failure, wait 3 seconds and retry once; after 3 consecutive failures across any combination of calls, stop and report. Never silently skip a failed search, and never present partial coverage as complete. **5. Require a retrievable link for every citation** No URL or DOI from this session means not citable. Enforce this at write time rather than at review time — it is the single check that catches the most problems. ### The audit log End any workflow that produces a deliverable with a table of what it actually did: | Search | Query | Filters used | Found | Returned | Cited | Status | | - | - | - | - | - | - | - | | 1 | creatine muscle hypertrophy | `human`, `year_min: 2015` | 20 | 10 | 4 | Success | | 2 | systematic review creatine strength | `study_types: systematic review` | 8 | 8 | 3 | Success | | 3 | creatine cognitive function | `human` | 0 | 0 | 0 | Success, no results | | 4 | creatine dosage timing | none | 11 | 11 | 2 | Failed once, retry succeeded | Follow it with the totals, any plan cap you detected, and any search that failed after its retry. This is what lets a reader — or a reviewer, or a mentor — judge how much weight the output deserves, and it is cheap to produce if you have been counting from the start. > [!TIP] > > A run that found nothing is a result worth recording. An empty audit log entry distinguishes "no new evidence this week" from "the job broke silently three weeks ago." ## What makes a good Consensus use case > [accordion] The question has an evidence base > > Consensus searches peer-reviewed literature. It answers "what does the published evidence say about X" extremely well. It is not a source for your internal experiment data, patent full text, regulatory filings, or private clinical data — join those to Consensus results rather than expecting Consensus to hold them. > [accordion] The population can be narrowed with filters > > The filters are where accuracy comes from. A question you can express as "human RCTs, 100+ participants, Q1 journals, since 2020" comes back far cleaner than an unfiltered keyword search. If you cannot describe the study population you want, spend a turn deciding it before you search. > [accordion] The output is verifiable > > Every result carries `title`, `authors`, `journal_name`, `publish_year`, `citation_count`, `doi`, and a `url` back to the paper on Consensus. Good use cases end in something a human can audit line by line — an evidence table, a cited memo, a screening decision log. Ones that end in an uncited assertion waste the citation trail. > [accordion] It decomposes into many searches > > Nearly every high-value workflow in the gallery is a fan-out. If your design issues one search and stops, you are leaving most of the quality on the table. Breadth comes from the number of queries, not from how fast you fire them — pace them to your tier's limit. > [accordion] Poor fit: single-fact lookups and non-research questions > > If the answer is one number from a database, a definition, or today's news, Consensus is the wrong tool. Reach for it when the answer is *contested*, *cumulative*, or *needs a citation*. ## Surface differences worth knowing | | MCP | REST API | | - | - | - | | Auth | OAuth per user, so each person's own plan and limits apply | One `x-api-key` for the whole integration | | Best for | Interactive work and agent frameworks; no code to write | Pipelines, batch sweeps, and anything scheduled | | `clinical_guideline` | Not exposed | Available | | `include_semantic_score` | Not exposed | Available | | Pagination control | Available on paid plans | Available on paid plans and Enterprise keys | > [card] [Use case library](/use-cases/gallery) > > Browse all 30 worked examples by persona. > [card] [Search endpoint reference](/api-reference/search) > > Every filter, the full response schema, and an interactive playground. # Create a biomarker evidence report Source: https://docs.consensus.app/use-cases/biomarker-report Assess the evidence for a biomarker's optimal range across Mendelian randomization, large cohorts, and trials — or establish that no evidence-based range exists. ## What you get A tiered evidence report on one biomarker: what range the published evidence actually supports for the outcomes you care about, which evidence tiers support it, where the tiers disagree, and — when the evidence cannot bound a range — an explicit statement that it cannot, instead of a plausible-sounding number. ## Who it's for Biotech and pharma teams, longevity clinicians, and translational researchers who need more than a lab reference interval. This is one of the heaviest real usage patterns on Consensus. ## The prompt The structure below is what makes this work: the definitional constraints and the per-tier instructions do more for output quality than any amount of prompt polish. Replace `[BIOMARKER]` and the outcome set. ```text theme={null} You are a clinical epidemiologist, longevity medicine researcher, and evidence synthesis expert. Task: Assess all available evidence to determine the evidence-based optimal range of the biomarker [BIOMARKER] in adults, optimized for longevity and disease-risk reduction, not merely the conventional lab reference interval. Population: Adults, separated by sex where biologically or clinically relevant. Note differences by age, menopausal status, pregnancy status, ethnicity, kidney function, body composition, medication use, assay type, and disease state when relevant. Core question: What biomarker range is most consistent with the lowest risk of all-cause mortality, disease-specific mortality, major age-related disease, frailty, and biological aging, where both the lower and upper bounds of the range are supported by evidence that values outside the range are less favorable? If such a true optimal range cannot be established, report the best-supported risk thresholds and explicitly state that no evidence-based optimal range is known. Critical definition of "optimal range": Do not define the optimal range as merely: - within the standard reference interval; - not clinically deficient; - not clinically excessive; - a range where higher or lower values have not been proven harmful; - a range used as a treatment target in a clinical trial; - a common clinical target; - the middle of the reference interval. Only call a range "optimal" if the evidence supports both its lower and upper bounds. If proposing a bound that lies inside the conventional reference interval, justify it with evidence that values beyond it are less favorable. Explicitly report whether risk rises on both sides of the proposed range. If only one side of the curve shows increased risk, do not claim a bounded optimal range unless other evidence supports the opposite boundary. Search each evidence tier separately using Consensus, and report them separately: A. Randomized controlled trials - Interventions that moved the biomarker and measured hard outcomes. - Note whether the trial targeted a range or a threshold. B. Systematic reviews and meta-analyses - Pooled dose-response relationships where available. C. Genetic and Mendelian randomization studies - Assess whether genetically predicted differences in the biomarker are associated with longevity, mortality, or disease outcomes. - Discuss pleiotropy, instrument strength, directionality, and whether MR supports causality. - State which MR supports: lower is better, higher is better, mid-range optimal, harm at high levels, harm at low levels, or no causal effect. - Distinguish causal evidence for the biomarker itself from causal evidence for related pathways or binding proteins. D. Large observational cohorts - Include dose-response, nonlinear, U-shaped, J-shaped, reverse J-shaped, or threshold analyses. - Extract nadir risk ranges where reported. - Note adjustment covariates, reverse causation risk, and residual confounding. - Report whether results differ after excluding early deaths, current smokers, people with prevalent disease, and people with frailty, inflammation, kidney disease, malnutrition, or major comorbidity. - Do not interpret absence of additional benefit at high or low values as evidence of harm. Output: 1. Proposed optimal range with both bounds, or an explicit statement that none can be established. 2. A table by evidence tier: what it shows, its direction, and its strength. 3. Where the tiers disagree, and the most likely reason. 4. Stratification notes where the answer differs by subgroup. 5. What evidence would most change this conclusion. Cite every claim with journal, year, and DOI. ``` ## How it works **1. The definitional block does the heavy lifting** Without the "do not define optimal range as…" list, models converge on the reference interval and call it an answer. Enumerating the wrong definitions is what forces a real evidence assessment. **2. Each tier gets its own search and its own instructions** Mendelian randomization and observational cohorts fail in different ways, so they need different critical questions. Searching them separately also makes tier disagreement visible instead of averaging it away. **3. Both bounds must be evidenced** This is the constraint that prevents a one-sided risk curve being reported as a bounded range — the most common error in biomarker summaries. **4. Negative results are a first-class output** "No evidence-based optimal range is known" is frequently the correct answer. Naming it up front stops the model manufacturing a range to satisfy the request. ## The API equivalent Run one call per tier and keep the results separate: ```bash theme={null} # Tier C — Mendelian randomization curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=homocysteine Mendelian randomization all-cause mortality causal" \ --data-urlencode "human=true" \ --data-urlencode "domain=med,bio" \ --data-urlencode "page_size=50" # Tier D — large cohorts, nonlinear dose-response curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=homocysteine U-shaped association all-cause mortality cohort nadir" \ --data-urlencode "study_types=cohort study,longitudinal / panel data study" \ --data-urlencode "human=true" \ --data-urlencode "sample_size_min=1000" \ --data-urlencode "page_size=50" ``` `medical_mode=true` narrows to top medical journals and guidelines (about 8M documents) when a tier returns too much general literature. ## What to check before you trust it > [!WARNING] > > This is a research summary for qualified human review, not clinical guidance. Nothing here should drive a treatment decision on its own. Keep the citation trail intact so a reviewer can check each statement against its source. * **Check that both bounds are actually evidenced.** The most common failure is a lower bound with real support and an upper bound asserted from convention. Read the justification for each separately. * **Reverse causation is the main confounder.** Low values of many biomarkers mark existing illness. If the report does not discuss the sensitivity analyses excluding early deaths and prevalent disease, the U-shape may be an artefact. * **Check assay comparability.** Ranges from different assays are often not interchangeable — the prompt asks for this, but verify it was answered. * **Watch for tier disagreement being smoothed over.** If MR says no causal effect and cohorts show a strong association, that conflict is the finding. ## Related > [card] [Build a drug target validation dossier](/use-cases/target-evidence-tiers) > > The same tiered method, applied to a target rather than a biomarker. > [card] [Extract numbers and formulas from full-text papers](/use-cases/extract-values-from-full-text) > > Pull the nadir ranges and effect sizes into rows you can compute on. > [card] [Best practices](/use-cases/best-practices) > > Fan-out, the filter ladder, and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Monitor new research about a drug, device, or brand Source: https://docs.consensus.app/use-cases/brand-and-device-monitoring Track what gets published about a drug, device, or brand, and route only the papers that change something. ## What you get A recurring brief covering what appeared in the literature about your products, your competitors' products, and the mechanisms behind them — filtered down to the items that matter, with an explicit empty state when nothing does. ## Who it's for Commercial, medical affairs, and competitive-intelligence teams in life sciences, and consultancies running this on behalf of clients. Also the shape used by companies building literature monitoring into their own product. ## How it works **1. Build the watchlist from mechanisms, not just brand names** Brand names are unreliable in the literature — papers name the molecule, the device class, or the mechanism. Query all of them: brand, generic or molecule name, device category, and the underlying mechanism. **2. Add the competitor and author axes** Once you know which authors publish in the space, track their names as queries in their own right. Shifts in who publishes often precede shifts in what is published. **3. Window and diff** Set `year_min` and `month_min` to the last run. Diff on `doi` against stored state so the brief contains only what is new. **4. Interpret, do not list** For each new paper, say what it implies — a competitor's program stage, a safety question, a comparator result. Flag changes in venue, geography, or collaborator. **5. Escalate on signal type, not volume** Route safety findings and head-to-head comparisons to a human immediately. Everything else can wait for the weekly brief. ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=drug-eluting stent restenosis outcomes comparison" \ --data-urlencode "year_min=2026" \ --data-urlencode "month_min=8" \ --data-urlencode "exclude_preprints=false" \ --data-urlencode "page_size=100" ``` Useful narrowing for this workflow: | Parameter | Why it helps here | | - | - | | `medical_mode=true` | Restricts to top medical journals and guidelines when general literature is too noisy | | `country` | Region-specific briefs for local commercial teams | | `journal_name` | Watch a specific venue where your category publishes | | `sjr_max=1` | Top-quartile only, when volume is high and you need the papers that will be cited | | `exclude_preprints=false` | Keep preprints — early signals often surface there first | ## Turning it into a product Several customers resell this. If that is the shape you are building: * **The API is per-integration, not per-user.** One `x-api-key` covers your service; your own users never need Consensus accounts. * **Size the plan against request volume.** A watchlist of 50 entities refreshed weekly, paginated, runs into the thousands of requests a month. * **Cache on the full parameter set.** Two clients watching overlapping mechanisms will issue near-identical queries. * **Pass the `url` through.** Every result links back to the paper on Consensus, which is the citation trail your users need. ## What to check before you trust it * **Brand-name-only watchlists miss most of the literature.** Verify by searching the molecule and mechanism separately and comparing volume. * **Alert when a run returns nothing across the whole watchlist.** That is usually a broken job, not a quiet week. * **Short brand names collide.** Spot-check any entity producing unusually high volume for unrelated hits. * **The brief must have an empty state and use it.** A monitor that always finds five things trains people to ignore it. ## Related > [card] [Track new papers across many topics on a schedule](/use-cases/scheduled-research-pipeline) > > The engineering underneath: pagination, diffing, backoff. > [card] [Build a drug target validation dossier](/use-cases/target-evidence-tiers) > > Go deep on a mechanism the monitor surfaced. > [card] [Best practices](/use-cases/best-practices) > > Diff-against-last-run and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Verify that every citation is real and supports its claim Source: https://docs.consensus.app/use-cases/citation-verification Verify every reference in a document actually exists, then confirm each one says what the text citing it claims. ## What you get A per-reference verdict: verified with a DOI, metadata mismatch, or not found. Then, for the ones that exist, a second verdict on whether the paper actually supports the sentence citing it. Two different failures, caught separately. A fabricated reference is obvious once you look. A real paper cited for something it never claimed is the one that survives review. ## Who it's for Anyone reviewing AI-assisted drafts, grant text, vendor reports, or student submissions. Also worth wiring into a submission pipeline if you handle volume. ## The prompt ```text theme={null} You are a reference-checking specialist. You are auditing a bibliography for fabricated and misapplied citations using Consensus. Stage 1 — does the reference exist? For each reference in the list below: 1. Search Consensus for the exact title, verbatim, in quotes. 2. Compare what comes back against the reference as given, field by field: title, first author surname, journal, and publication year. 3. Assign a status: - VERIFIED — title matches near-exactly and the first author, journal, and year all agree. Record the DOI you matched. - METADATA MISMATCH — a paper with this title exists, but one or more of author, journal, or year is wrong. Say which field and give the correct value. - NOT FOUND — no near-title match. Before concluding this, retry once with the distinctive noun phrase from the title rather than the full string. Do not mark a reference VERIFIED on topical similarity. A different paper about the same subject is NOT a match — that is the exact failure being hunted here. Stage 2 — does it support the sentence? For each VERIFIED reference, take the sentence in the document that cites it and search Consensus for that claim. Then judge: - SUPPORTS — the paper substantiates the sentence. Quote the supporting text. - DOES NOT SUPPORT — the paper exists and is on-topic but does not make this claim. Say what it does claim instead. - CANNOT TELL — the abstract is insufficient to judge. Say what you would need to read. Output two tables, one per stage. End with counts: verified, mismatched, not found, and how many verified references fail to support their sentence. References: [PASTE THE BIBLIOGRAPHY] Document text (for stage 2): [PASTE THE TEXT, or omit to run stage 1 only] ``` ## How it works **1. Exact title search separates real from invented** A real paper returns a near-exact `title` match. A fabricated one returns only loose topical neighbours — which is precisely why the prompt forbids accepting topical similarity as a match. **2. Metadata is the tiebreak, not judgment** Requiring `publish_year`, `journal_name`, and first author to agree before marking anything verified turns a soft call into a mechanical one. Hallucinated references very often pair a real title with a wrong year or journal. **3. One retry on the distinctive phrase** Long titles with subtitles sometimes miss on the full string. A single retry on the distinctive noun phrase prevents false `NOT FOUND` verdicts without opening the door to loose matching. **4. Stage 2 is the one that finds real problems** In AI-assisted drafts, most references exist. The failure that matters is a genuine paper attached to a claim it never made. ## The API equivalent Stage 1, per reference — search the title and compare metadata: ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=Effects of intermittent fasting on cardiometabolic risk factors" \ --data-urlencode "page_size=5" ``` Match on a normalized `title` comparison, then assert `publish_year`, `journal_name`, and the first entry of `authors`. Record `doi` on success. Stage 2 — search the citing claim and read the excerpt: ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=intermittent fasting LDL cholesterol reduction magnitude" \ --data-urlencode "page_size=10" ``` > [!NOTE] > > Full-text excerpts (`full_text_chunks`) come back on paid plans and cover open-access papers only. Stage 1 needs none of it: title and metadata matching works on every plan. ## What to check before you trust it * **A `NOT FOUND` is not proof of fabrication.** Book chapters, conference proceedings, grey literature, and very recent papers may sit outside coverage. Treat it as "needs a human look", not a verdict. * **Beware near-identical titles.** Some fields have several papers with nearly the same title across different years. The author and year assertions are what keep these apart. * **Run stage 1 alone first.** It is cheap, needs no full text, and usually surfaces the worst problems on its own. * **Do not auto-reject on this output.** It is a triage tool that tells a human where to look. ## Related > [card] [Find a real citation for every claim in a draft](/use-cases/ground-a-manuscript) > > The upstream fix: get real citations attached as you write. > [card] [Substantiate a product claim with published evidence](/use-cases/claim-substantiation) > > The same verification discipline, applied to claims that ship. > [card] [Best practices](/use-cases/best-practices) > > Fan-out, the filter ladder, and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Substantiate a product claim with published evidence Source: https://docs.consensus.app/use-cases/claim-substantiation Build the evidence file behind a claim before it ships — including what contradicts it. ## What you get A claim-by-claim substantiation file: the peer-reviewed evidence supporting each claim, the evidence against it, a defensibility grade, and suggested wording the evidence actually supports. ## Who it's for Product, brand, regulatory, and R\&D teams in consumer goods, sports and performance products, supplements, and devices — plus the consultancies who advise them. ## The prompt ```text theme={null} You are a claims substantiation analyst. You are assembling the evidence file behind product claims before they go to market. Claims: [LIST THE CLAIMS, ONE PER LINE] Product context: [WHAT IT IS, WHO USES IT, HOW] For each claim: 1. Split it into testable parts. "Reduces injury risk and improves comfort" is two claims with different evidence bases — never assess them together. 2. State each part as a proposition naming the population, the intervention or product attribute, the comparator, and the outcome measure. 3. Search Consensus for supporting evidence. Require human studies with a control condition and a realistic sample size. Note the exact population studied — trained athletes are not recreational users, and lab protocols are not real-world use. 4. Search separately for contradicting evidence, using the negative phrasing. Finding the contradicting paper yourself is far cheaper than a competitor or a regulator finding it. 5. Check whether the effect was measured directly or inferred from a surrogate. A claim about injury risk supported only by a biomechanical proxy is a different claim. Grade each part: STRONG — multiple controlled human studies, consistent direction, population matches our users. QUALIFIED — evidence exists but is limited by population, sample size, design, or reliance on a surrogate outcome. Write the narrower claim it supports. UNSUPPORTED — no adequate evidence found. Say what you searched. CONTRADICTED — the weight of evidence points the other way. Give the citation. For every QUALIFIED and UNSUPPORTED part, propose the wording the evidence does support, and say what study would be needed to support the original. Output: Claim part | Grade | Population studied | Direct or surrogate outcome | Supporting citations | Contradicting citations | Suggested wording. ``` ## The API call Demand controlled human evidence at a realistic size: ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=forefoot bending stiffness running economy oxygen cost" \ --data-urlencode "human=true" \ --data-urlencode "controlled=true" \ --data-urlencode "sample_size_min=15" \ --data-urlencode "study_types=rct,non-randomized experimental study" \ --data-urlencode "page_size=50" ``` Then run the negative pass: ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=footwear stiffness no effect running economy null finding" \ --data-urlencode "human=true" \ --data-urlencode "controlled=true" \ --data-urlencode "page_size=50" ``` | Parameter | Why it matters here | | - | - | | `human=true` | Animal and in vitro evidence rarely substantiates a consumer claim | | `controlled=true` | Uncontrolled studies will not survive review | | `sample_size_min` | Filters the underpowered studies that make weak claims look supported | | `sjr_max` | Restrict to higher-quartile journals when the field has a large promotional literature | | `domain` | Keeps technical terms in the right field | ## What to check before you trust it > [!WARNING] > > This is an evidence file for legal, regulatory, and marketing review — not a compliance sign-off. Requirements differ by market and claim type, and a human with that jurisdiction's expertise makes the final call. * **Check the population column on every row.** The most common substantiation failure is real evidence from a population that is not your customer. * **Direct versus surrogate outcome is load-bearing.** A biomechanical proxy does not substantiate an injury-reduction claim, however strong the correlation. * **Confirm the negative pass actually ran.** If there are no contradicting citations anywhere in the file, it was skipped — null findings exist in almost every applied field. * **Take the suggested wording seriously.** The narrower claim is usually still a good claim, and it is the one you can defend. ## Related > [card] [Extract numbers and formulas from full-text papers](/use-cases/extract-values-from-full-text) > > Pull the effect sizes behind the claim into a comparable table. > [card] [Find human-factors evidence for a design decision](/use-cases/human-factors-evidence) > > The human-subjects evidence behind a product decision. > [card] [Best practices](/use-cases/best-practices) > > The filter ladder and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Grade a set of papers with your own evidence rubric Source: https://docs.consensus.app/use-cases/evidence-grading-rubric Apply your organization's standard for what counts as good evidence, consistently, across every paper on a topic. ## What you get Every paper scored against the same rubric, with the reason for each score and the evidence behind it — so two people asking the same question a month apart get comparable answers. This is what our highest-volume self-serve user does: a biotech founder running a custom evidence-grading prompt over hundreds of searches. ## Who it's for Any team with an internal standard for evidence quality — biotech, clinical, nutrition, or research operations — where "is this well-supported?" currently depends on who you ask. ## The prompt ```text theme={null} You are an evidence assessor applying a fixed grading rubric. Question: [THE CLAIM OR TOPIC] Rubric (score every paper on all three axes): A. Design strength 5 — meta-analysis or systematic review of controlled trials 4 — randomized controlled trial 3 — prospective cohort or well-controlled non-randomized study 2 — cross-sectional, case-control, or retrospective 1 — case report, in vitro, or animal only B. Execution quality 5 — adequately powered, pre-registered, low attrition, blinded where possible 3 — reasonable but with named limitations 1 — serious limitations: underpowered, high attrition, no control C. Relevance to our question 5 — same population, same exposure, same outcome 3 — differs on one dimension; say which 1 — indirect or surrogate outcome only Rules: - Score from what the paper reports, not from journal or author reputation. - Where a paper does not report enough to score an axis, mark it NOT REPORTED rather than assuming the middle. Do not average around missing information. - A high design score does not raise the execution score. A large RCT executed badly is a 4 and a 1, and that combination is the useful signal. - State the single fact that most constrains each score. Output: Paper | DOI | A | B | C | Overall | Limiting factor | Supporting quote. Then: the highest-graded evidence, the strongest contradicting evidence, and what an ideal study would need to settle the question. ``` ## Getting a rubric worth applying **1. Separate the axes** Design, execution, and relevance fail independently. A single 1–5 "quality" score hides which one is the problem and makes the grades incomparable across topics. **2. Anchor each level with an example** "Well-controlled" means different things to different readers. Naming a real paper at each level makes the scale reproducible. **3. Prohibit prestige as a proxy** Journal and author reputation are already available as `sjr_best_quartile` and `citation_count`. Keep them as separate metadata, not smuggled into the quality score. **4. Make NOT REPORTED a real value** Papers that omit a detail are common. Scoring them at the midpoint quietly inflates the evidence base. ## Feeding it good candidates ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=omega-3 supplementation cardiovascular events" \ --data-urlencode "study_types=meta-analysis,systematic review,rct" \ --data-urlencode "human=true" \ --data-urlencode "sample_size_min=100" \ --data-urlencode "page_size=50" ``` Paid plans return `study_type` and `takeaway` on each result, which give the grader a head start on axis A. Full-text excerpts (`full_text_chunks`, paid plans) are what make axis B possible at all: execution quality is rarely visible in an abstract. ## Audit requirements Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability). Record the rubric version alongside every graded set: a rubric edit makes old grades incomparable, and teams discover this months later when two reports disagree. ## What to check before you trust it * **Check the score distribution.** If nothing scores low, the rubric is being applied generously — real literature has a spread. * **Grade five papers by hand first** and compare. Disagreements almost always point at an ambiguous rubric level, not a model failure. * **Watch for axis contamination.** If design and execution scores move together on every paper, they are not being assessed separately. * **Execution scores from abstracts are unreliable.** Without full text, mark axis B as low-confidence rather than guessing. ## Related > [card] [Create a biomarker evidence report](/use-cases/biomarker-report) > > Tiered grading applied to one biomarker question. > [card] [Check whether a finding has human evidence yet](/use-cases/translational-readiness) > > Split preclinical from human evidence before grading it. > [card] [Best practices](/use-cases/best-practices) > > Grounding rules and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Extract numbers and formulas from full-text papers Source: https://docs.consensus.app/use-cases/extract-values-from-full-text Pull numbers, formulas, and parameters out of papers with their measurement conditions attached. ## What you get Rows you can compute on: one row per extracted value, each carrying the conditions it was measured under and the excerpt it came from. Not a summary of papers — a dataset built from them. This is what teams reach for when abstracts are not enough: an aerodynamics group extracting mathematical formulas from full texts, a materials team collecting performance figures across test conditions, a clinical team pulling effect sizes for a meta-analysis. ## Who it's for Engineering and physical-science R\&D, and any research team assembling a quantitative dataset from published work. > [!NOTE] > > Full-text excerpts (`full_text_chunks`) come back on paid plans and cover open-access papers only, up to the first 20 on each page. Everything else on this page (search, filtering, metadata) works on any plan. ## How it works **1. Fix the schema before extracting** Decide every column up front: the value, its units, the conditions it was measured under, the method, the sample or specimen, and the source. A schema settled mid-run produces a table you cannot pivot. **2. Phrase the query as the thing being extracted** Excerpts are selected for relevance to your `query`. Searching the paper's topic returns introduction paragraphs; searching the measurement returns the results section. This single choice does more for extraction quality than anything else. **3. Set the domain, or drown** Without `domain`, technical terms pull in unrelated biomedical hits. `domain=mat,eng,phys,chem` is the difference between a usable and an unusable result set. **4. Keep every value with its conditions and its excerpt** A number without its test conditions is not comparable to another number. A number without its excerpt is not auditable. Both travel with the row. **5. Distinguish not-reported from not-extracted** Use different values for the two. Conflating them silently biases every downstream count. ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=turbulent boundary layer separation drag coefficient measured values" \ --data-urlencode "domain=eng,phys,mat" \ --data-urlencode "page_size=100" ``` Each result's `full_text_chunks` holds query-relevant excerpts. Pair with metadata already on the result — `doi`, `title`, `publish_year`, `journal_name`, `url` — so each extracted row carries its provenance. ## The extraction prompt ```text theme={null} You are a data extraction specialist building a structured dataset from papers. Schema — every row must have all of these, or an explicit marker for why not: value | units | quantity measured | measurement conditions | method or apparatus | specimen or sample | source DOI | source excerpt Rules: 1. Extract only values that appear in the excerpt you were given. Never infer, convert, or interpolate a value that is not written there. 2. Every value keeps the conditions it was measured under. If the excerpt gives a number without its conditions, record it as CONDITIONS NOT REPORTED — do not assume standard conditions. 3. Preserve the original units. Add a converted column if needed, but never overwrite the original. 4. Use NOT REPORTED when the paper does not give a field, and NOT EXTRACTED when the excerpt was insufficient. These are different and must not be merged. 5. Copy the source excerpt verbatim into its column. Every row must be checkable against it without reopening the paper. 6. For formulas, reproduce the expression exactly and define every symbol as the paper defines it. If a symbol is defined elsewhere in the paper and not in your excerpt, mark it undefined rather than guessing. Output the rows as CSV. Follow it with a list of papers you could not extract from and why. ``` ## What to check before you trust it * **Spot-check five rows against their excerpts.** The excerpt column exists precisely so this takes two minutes rather than an afternoon. * **Watch for silent unit assumptions.** Values converted without an original-units column are the most common corruption in extracted datasets. * **Check the CONDITIONS NOT REPORTED rate.** If it is near zero, conditions are being assumed rather than read. * **Coverage is not uniform.** Full-text excerpts currently cover open-access papers, so a closed-access subfield will be under-represented. Check whether your topic skews that way before treating the dataset as complete. ## Related > [card] [Substantiate a product claim with published evidence](/use-cases/claim-substantiation) > > Turn extracted evidence into a defensible claim file. > [card] [Create a biomarker evidence report](/use-cases/biomarker-report) > > The same extraction discipline inside a clinical evidence review. > [card] [Best practices](/use-cases/best-practices) > > Excerpt-first extraction and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Find published work on a failure or degradation mechanism Source: https://docs.consensus.app/use-cases/failure-mode-scan Check whether a defect you are seeing is a known mechanism with a published cause, before you spend a quarter rediscovering it. ## What you get What the literature already reports about your failure mode: proposed mechanisms, the conditions that drive it, published mitigations, and how others detected it — with the citations to hand to your reliability team. ## Who it's for Reliability engineers, failure analysis teams, and process integration groups in semiconductor, automotive, and aerospace manufacturing. ## Why check the literature first Failure and degradation mechanisms are heavily studied academically, often more thoroughly than any single company can afford internally. Checking whether a mechanism is known costs an afternoon; rediscovering it costs a quarter. The literature also gives you the vocabulary — knowing the accepted name for what you are seeing makes every subsequent search work. ## How it works **1. Describe the observation, not your hypothesis** Search what you actually see — the morphology, where it appears, what conditions precede it. Searching your suspected cause finds papers confirming that cause, whether or not it is right. **2. Search the mechanism vocabulary separately** Once results suggest a named mechanism, search that name directly. Field-standard terminology returns far better results than descriptive language, and you rarely know it at the start. **3. Cover the adjacent industries** The same physical mechanism often appears under a different name in another field. Electromigration, fretting, delamination, and thermal fatigue all cross industries. Widen `domain` deliberately once before narrowing. **4. Search for mitigations and for failed mitigations** Both matter. A published mitigation that did not work saves you from trying it. **5. Pull detection and characterization methods** How other groups measured the mechanism tells you what to instrument, and is often the fastest thing to act on. ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=electromigration void nucleation copper interconnect current density" \ --data-urlencode "domain=eng,phys,mat" \ --data-urlencode "page_size=50" ``` Add `year_min` when the relevant process generation is recent — failure behaviour is often node- or generation-specific, and older literature may describe a materially different structure. ## What to check before you trust it * **Confirm the geometry and scale match.** A mechanism documented at one dimension or process node may behave differently at another, and papers do not always make the regime explicit. * **Check the accelerated-test conditions.** Most reliability literature uses accelerated ageing. The extrapolation back to operating conditions is where the assumptions hide. * **A mechanism that matches your symptoms is a hypothesis.** Several mechanisms produce similar signatures; use the literature to design the discriminating test, not to close the investigation. * **Coverage skews toward open access.** Much of this literature sits in closed conference proceedings, so treat a thin result as inconclusive rather than as evidence the mechanism is unstudied. ## Related > [card] [Compare candidate materials against published test data](/use-cases/materials-comparison) > > Once you know the mechanism, compare materials that resist it. > [card] [Find published process conditions for a manufacturing step](/use-cases/process-window-precedent) > > Conditions that avoid the mechanism entirely. > [card] [Best practices](/use-cases/best-practices) > > Fan-out and merge, and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Use case library Source: https://docs.consensus.app/use-cases/gallery Worked examples of what people build on Consensus, filterable by persona, each with the full prompt and API call behind it. Browse what researchers, life-science teams, engineers, and platform builders actually build on Consensus. Every card links to the full prompt or API call you can copy and run today. New here? Start with the [use cases overview](/use-cases). Who you are All Academic Life & health sciences Applied & physical sciences Platform Showing all 31 use cases ★ Academic Find a real citation for every claim in a draft Paste a draft section and get every claim matched to a real, verifiable paper — or flagged as unsupported. MCP [Open](/use-cases/ground-a-manuscript) ★ Academic · Platform Verify that every citation is real and supports its claim Verify every reference actually exists, then confirm it says what the text citing it claims. MCP API [Open](/use-cases/citation-verification) ★ Platform · Academic Build a reusable literature-review workflow Wrap a repeatable review workflow into a skill your whole team can run the same way every time. MCP API [Open](/use-cases/literature-review-skill) Platform · Academic Track new papers across many topics on a schedule Sweep hundreds of topics on a cron, diff against last run, and report only what changed. API Scheduled [Open](/use-cases/scheduled-research-pipeline) Life & health sciences Create a biomarker evidence report Assess the evidence for a biomarker's optimal range across Mendelian randomization, cohorts, and trials. MCP API [Open](/use-cases/biomarker-report) Life & health sciences Build a drug target validation dossier Human genetics, in vitro, animal, and clinical evidence for one target — including what contradicts it. API Fan-out [Open](/use-cases/target-evidence-tiers) Life & health sciences Monitor new research about a drug, device, or brand Track what gets published about a drug, device, or brand and route only the consequential papers. API Scheduled [Open](/use-cases/brand-and-device-monitoring) Academic Turn a syllabus into a recommended reading list Extract course topics and return recent papers mapped to each learning outcome. MCP [Open](/use-cases/syllabus-reading-list) Academic Build and maintain a Library collection from your assistant Find papers, save them to a collection, then list, filter, and rename it without opening Consensus. MCP API [Open](/use-cases/library-reading-list) Academic · Life sciences Find NIH grants for a research idea Position your idea, map it to the right institutes and study sections, and surface NOSIs and funded overlap. MCP API [Open](/use-cases/grant-finder) Applied & physical sciences · Academic Extract numbers and formulas from full-text papers Pull numbers, formulas, and parameters out of papers with their measurement conditions attached. API Full text [Open](/use-cases/extract-values-from-full-text) Applied & physical sciences Substantiate a product claim with published evidence Build the evidence file behind a claim before it ships, including what contradicts it. API Filters [Open](/use-cases/claim-substantiation) ★ Platform Swap a hand-maintained paper corpus for full-text search Replace a curated PubMed subset you maintain yourself with a weekly-refreshed full-text index. MCP API [Open](/use-cases/swap-a-paper-corpus) ★ Platform Add a citable evidence layer to an agent you already ship Give a working agent access to peer-reviewed literature as a tool, so its claims carry citations. MCP API [Open](/use-cases/add-an-evidence-layer) Platform Answer one question with many parallel searches Decompose a question into search angles, gather across all of them, and distill one cited answer. API Fan-out [Open](/use-cases/parallel-search-angles) Platform Screen thousands of target-disease pairs for evidence Pull literature evidence for every pair in a large matrix — the batch workload that breaks most APIs. API Bulk [Open](/use-cases/high-throughput-screening) Platform Benchmark literature providers against your own ground truth Curate queries with known-relevant papers and measure which provider actually surfaces them. API [Open](/use-cases/benchmark-retrieval) Academic Screen abstracts against inclusion criteria Apply explicit criteria across thousands of abstracts, with a decision log a second reviewer can audit. API Filters [Open](/use-cases/abstract-screening) Academic · Life & health sciences Build a PRISMA-ready search corpus Assemble a reproducible, deduplicated record set with the flow counts a systematic review must report. API Bulk [Open](/use-cases/prisma-corpus) Academic Write the search-strategy section of your methods Turn a logged search into the reproducible methods paragraph reviewers ask for. API [Open](/use-cases/search-strategy-methods) Academic Find which journals to submit to Identify the journals actually publishing work like yours, with their quartile, so you target realistically. API [Open](/use-cases/where-to-submit) Academic Track how a field's terminology changed over time Compare old and recent literature so you search the terms the field actually used, not just today's. API Filters [Open](/use-cases/terminology-shift) Life & health sciences Grade a set of papers with your own evidence rubric Apply your organization's standard for good evidence consistently across every paper on a topic. MCP Full text [Open](/use-cases/evidence-grading-rubric) Life & health sciences Find the leading investigators in an indication Rank the authors publishing most in a therapeutic area, split by region and by recent versus lifetime output. API Bulk [Open](/use-cases/investigator-mapping) Life & health sciences Assemble a payer evidence dossier Pull cost-effectiveness, burden-of-illness, and real-world evidence a reimbursement submission needs. API Fan-out [Open](/use-cases/payer-evidence-dossier) Life & health sciences Check whether a finding has human evidence yet Separate cell and animal work from human studies, and report the ratio. API Filters [Open](/use-cases/translational-readiness) Applied & physical sciences Compare candidate materials against published test data Search performance data for each candidate under your operating conditions, and find the untested gaps. API Full text [Open](/use-cases/materials-comparison) Applied & physical sciences Find published work on a failure or degradation mechanism Check whether a defect you are seeing is a known mechanism before you spend a quarter rediscovering it. API Fan-out [Open](/use-cases/failure-mode-scan) Applied & physical sciences Find published process conditions for a manufacturing step See what temperature, time, and concentration ranges others published, so you start from precedent. API Full text [Open](/use-cases/process-window-precedent) Applied & physical sciences Check a supplier's performance claim against independent research Test what a vendor claims against what independent published work reports. API Filters [Open](/use-cases/supplier-claim-check) Applied & physical sciences · Life & health sciences Find human-factors evidence for a design decision Pull the biomechanics, ergonomics, and usability research behind a product or workstation choice. API Filters [Open](/use-cases/human-factors-evidence) No use cases match that persona yet. Pick **All** to see everything. ## Next steps > [card] [Build internal research tools](/use-cases/internal-tools) > > The guide for organizations: internal assistants, monitoring agents, literature tables, and agents. > [card] [Build custom research systems](/use-cases/research-systems) > > The guide for academics: literature reviews, grounded writing, and evidence reports. > [card] [Best practices](/use-cases/best-practices) > > Five primitives every workflow above is built from, and what makes a good Consensus use case. > [card] [Connect MCP](/consensus-mcp) > > Run any prompt on this page from Claude, ChatGPT, Claude Code, or Codex in under two minutes. > [card] [Get started with the API](/api-get-started) > > Your first authenticated request to `GET /v1/search`, in seven languages. > [card] [Talk to us about Enterprise](https://consensus.app/home/api) > > Higher page sizes, unlimited usage, and custom rate limits for corpus-scale work. # Start building with the Consensus API and MCP Source: https://docs.consensus.app/use-cases/get-started Pick a surface, run your first filtered search, and find the guide for what you are building. Four steps from zero to a working integration. **1. Pick your surface** Use **MCP** to try Consensus inside Claude or ChatGPT in two minutes, or when each person should search under their own account. Use the **API** for anything in code — monitoring agents, literature tables, pipelines, and scientific agents running on one shared key. **2. Run one filtered search** [Make your first request](/api-get-started), then add filters one at a time. The filters are where precision comes from. ```bash theme={null} curl -G "https://api.consensus.app/v1/search" -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=vision-language models for robot manipulation" \ --data-urlencode "domain=cs" \ --data-urlencode "year_min=2024" ``` **3. Read the guide for what you are building** Start with [internal tools](/use-cases/internal-tools) or [research systems](/use-cases/research-systems), then read [best practices](/use-cases/best-practices) for rate limits, pagination, and grounding rules. **4. Scale up when you need to** Enterprise adds up to 1,000 results per request, custom call volumes, and custom rate limits. [Talk to us](https://consensus.app/home/api) about a trial. ## Next > [card] [Build internal research tools](/use-cases/internal-tools) > > Assistants, field-monitoring agents, literature tables, and scientific agents. > [card] [Build custom research systems](/use-cases/research-systems) > > Literature reviews, grounded writing, and evidence reports. # Find NIH grants for a research idea Source: https://docs.consensus.app/use-cases/grant-finder Position a research idea against the literature, map it to the right NIH institutes and study sections, and surface NOSIs, open FOAs, and funded overlap. ## What you get An editable Word document scoping the NIH funding landscape for one research idea: draft Significance and Innovation language built from real gap quotes, a ranked list of institutes actually funding this work, the study sections that would review it, any active Notices of Special Interest, and mechanism recommendations matched to career stage *and* project scope. Consensus ships this as a packaged skill. It covers NIH only — not PCORI, DOD CDMRP, VA Merit, or foundations — and says so up front rather than implying broader coverage. ## Who it's for Clinical and translational researchers scoping funding for an idea, from trainees through established investigators. Also mentors and research-development offices advising them. ## Phase 1 — Intake Set expectations first, then ask three questions: 1. **Career stage** — trainee (F31/F32/T32), early-career faculty (K01/K08/K23/K99-R00/R21), or established. If established, split mid-career (has a K or first R01 → R01/R21/R34) from senior (R01/R35/P01/U01). 2. **Preliminary data** — starting from scratch, pilot evidence available, or resubmission. If resubmission, ask whether it was scored or triaged; that changes the strategic advice entirely. 3. **Institutional environment** — major research university, academic medical center, or another setting (smaller/community-based, or VA/military-affiliated). This affects which mechanisms are realistic and how to frame the Environment section. ## Phase 2A — Positioning analysis, five searches The goal is **not** to answer "is this novel?" It is to produce **positioning language the researcher can paste into a grant**. Each search builds a different part of the argument. Run them sequentially — see [rate limits](/use-cases/best-practices#rate-limits-and-result-caps). | # | Facet | What it produces | | - | - | - | | 1 | What's established | The "what is known" setup in Significance | | 2 | The problem's importance | Prevalence, burden, disparities — the "why now" | | 3 | Current approaches and limits | What exists today, and therefore what's insufficient | | 4 | The method in adjacent contexts | Applied elsewhere = validation; nowhere = novelty | | 5 | **What researchers say is missing** | The highest-value search — the gap, in their words | Search 5 matters most. Look explicitly for language like *"future research should"*, *"remains unclear"*, *"no studies have"*, *"limited evidence"*, *"critical gap"*. Pull **3–5 of these with full attribution** — they become the backbone of the positioning argument, because they let the researcher show reviewers that the field itself recognizes the gap. ### Build the narrative, don't list counts Deduplicate across all five facets by DOI first; papers found in multiple facets are high-signal. Then write 2–3 paragraphs the researcher can adapt directly: > **Paragraph 1** — "The field has established \[X, Y, Z from searches 1–2]. The burden is significant because \[stakes from search 2]." > > **Paragraph 2** — "Current approaches include \[search 3], but they fall short because \[gap quotes from search 5]. Specifically, \[quote 1] and \[quote 2]." > > **Paragraph 3** — "This project proposes \[the idea]. \[Search 4] suggests the approach is feasible, and the gap above shows it has not been applied to \[this population]. This is a \[novel/underexplored] intersection." **Evidence-depth gate:** aim for at least 3 directly relevant papers per facet and 15–20 unique cited papers overall — but only where the returned evidence supports it. Never add a marginal paper to pad the bibliography. Keep gap quotes under 25 words each and paraphrase the surrounding context. ## Phase 2B — Institute and grant discovery The NIH RePORTER API requires **POST**, so a GET-only fetch will not work: ```bash theme={null} CURRENT_YEAR=$(date +%Y) FY_RANGE="[$((CURRENT_YEAR-3)), $((CURRENT_YEAR-2)), $((CURRENT_YEAR-1)), $CURRENT_YEAR]" curl -s -X POST "https://api.reporter.nih.gov/v2/projects/search" \ -H "Content-Type: application/json" \ -d '{ "criteria": { "advanced_text_search": { "operator": "and", "search_field": "projecttitle,terms,abstract", "search_text": "NARROW KEYWORDS" }, "fiscal_years": '"$FY_RANGE"', "include_active_projects": true }, "limit": 20, "offset": 0, "include_fields": ["ProjectTitle","AbstractText","ActivityCode","AwardAmount","FiscalYear","PrincipalInvestigators","Organization","OpportunityNumber","AgencyIcAdmin","ProjectNum","StudySection"] }' ``` Run a narrow AND search and a broad OR search, then deduplicate by `project_num`. Compute the fiscal-year window dynamically so it never goes stale. **Search-quality gate.** Valid JSON is not a usable portfolio. If fewer than half the returned projects are substantively relevant, refine the keywords and re-run rather than ranking institutes off noise. If no focused search yields at least 5 relevant projects, proceed but label all institute conclusions preliminary. ### The most actionable finding Produce **two separate institute rankings** and explain any gap between them: 1. **Portfolio volume** — tally `agency_ic_admin.abbreviation`. 2. **Design and mission fit** — which institutes fund projects similar in population, methodology, clinical setting, and intended outcome. Do not simply recommend the biggest number. > "NIAID has the largest raw HSV portfolio, but most projects are basic virology or therapeutic development. NICHD has fewer awards but the closest methodological overlap with multicenter pediatric clinical prediction research; NICHD is therefore the stronger primary fit and NIAID a plausible secondary assignment." Then tally `study_section` values to identify the review panels, extract `opportunity_number` values starting with `NOT-` to find NOSIs, and match mechanisms on **both** career stage and project scope — a multi-site study does not fit in a 2-year \$275K R21, and a pilot does not need an R01. ## Phase 3 — The document Nine numbered sections, in order, none of which may be dropped for sparse data: Executive Summary · Your Positioning in the Field · Where NIH Is Funding This Work · Recommended Funding Opportunities · Similar Funded Research · Where This Work Gets Reviewed · Strategic Recommendations & Next Steps · References · Audit Log. If study-section data is unavailable, the section still appears and explains the limitation. Missing data never authorizes removing a required section or renumbering later ones. **Always include a program officer recommendation** — it is the single most valuable piece of advice for any applicant. POs confirm fit, suggest the right FOA, and flag problems before submission. Verify submission dates against a live NIH page rather than trusting a static table, and note which upcoming cycle is realistic given how much work remains. > [!WARNING] > > Budget ranges, submission dates, and eligibility change. Anything drawn from reference knowledge rather than a tool call in this session must be labelled as such and verified against the current official announcement before the researcher acts on it. ## Audit requirements Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability). This workflow needs two audit tables — Consensus searches (search #, facet, query, papers found, papers shown, papers cited, status) and RePORTER searches (type, keywords, total/returned, relevant projects used, status), including every refinement query. Add NOSI fetch outcomes where applicable. Distinguish three numbers explicitly: **found by Consensus**, **shown to us**, and **cited in this report**. If a plan cap was detected, say so — sparse results on a facet may be a plan ceiling rather than a genuine gap, and the researcher needs to know which. ## What to check before you trust it * **Check the gap quotes are real quotes**, with named attribution and a working link. Unexplained sentence fragments are the failure mode here. * **Read the institute reasoning, not just the ranking.** The volume-versus-fit distinction is the most actionable output; if the report just names the biggest institute, that analysis was skipped. * **Verify every FOA and NOSI** against its official NIH page before acting. Constructed links must not be presented as verified. * **A sparse RePORTER portfolio is information.** It may mean an unfunded niche — an opportunity or a warning, depending on why. ## Related > [card] [Build a reusable literature-review workflow](/use-cases/literature-review-skill) > > The deeper review workflow behind the positioning analysis. > [card] [Find a real citation for every claim in a draft](/use-cases/ground-a-manuscript) > > Once the grant is funded, cite the same way. > [card] [Best practices](/use-cases/best-practices) > > Rate limits, grounding rules, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Find a real citation for every claim in a draft Source: https://docs.consensus.app/use-cases/ground-a-manuscript Paste a draft section and get every empirical claim matched to a real, verifiable paper — or flagged as unsupported before a reviewer finds it. ## What you get Your draft back, claim by claim, with a real citation attached to each one — journal, year, DOI, and a link to the paper on Consensus. Claims the literature does not support come back flagged rather than quietly cited to something adjacent. This is the most common thing people do with Consensus MCP. Researchers writing with an AI assistant report going from citations they could not verify at all to a draft where nearly every claim checks out. ## Who it's for Anyone drafting with an LLM in the loop: PhD students writing a thesis chapter, clinicians writing a discussion section, postdocs turning results into a manuscript. Works in whatever you already write with — Claude, ChatGPT, Claude Code. ## The prompt Paste your draft section after this. Works as-is in any MCP client. ```text theme={null} You are a research librarian and evidence synthesis specialist. Task: For the draft text below, find a real peer-reviewed citation for every empirical claim, using Consensus. Do not write new prose. Do not improve my argument. Procedure, per claim: 1. Extract the claim as a standalone proposition. Name the population, the exposure or intervention, and the outcome. If any of those three is missing from my text, say so rather than assuming one. 2. Search Consensus for it. Run at least three phrasings covering synonyms and the terminology the relevant field actually uses. 3. Prefer the strongest available design: meta-analyses and systematic reviews first, then randomized trials, then observational studies. Note which tier you ended up in. 4. Read the abstract and the key takeaway. Confirm the paper supports the claim as I stated it, not merely the general topic. Classify every claim as exactly one of: - SUPPORTED — a paper supports it as stated. Give the citation and one quoted sentence or takeaway showing why. - PARTLY SUPPORTED — the evidence supports a narrower version. Give the citation and write the narrower claim I should use instead. - CONTRADICTED — the weight of evidence points the other way. Give the citation. - NO EVIDENCE FOUND — you searched and found nothing on point. Say what you searched. Do not substitute a loosely related paper. Rules: - Never cite a paper you did not retrieve through Consensus in this session. - Do not merge two claims to make one citation fit. - Common-knowledge and methodological statements need no citation. Skip them and say you did. Output a table: Claim | Verdict | Citation (authors, year, journal, DOI) | Evidence quote | Suggested rewording if needed. Draft text: [PASTE YOUR DRAFT SECTION] ``` ## How it works **1. Claims get isolated before anything is searched** Most bad citation matching happens because a sentence carrying three claims gets one citation. Forcing each claim into a standalone proposition with a named population, exposure, and outcome is what makes the search specific enough to be checkable. **2. Several phrasings per claim** A single phrasing is one search. Three or more across synonyms and field-specific terminology is the difference between "no evidence found" and finding the paper that exists. **3. Design tier is recorded, not just the citation** Knowing you ended up citing an observational study rather than a trial changes how you should word the sentence. Ask for the tier and you get that for free. **4. Four verdicts, not two** The `PARTLY SUPPORTED` bucket is where most of the value is — it turns an overclaim into a defensible sentence instead of forcing a binary keep-or-cut decision. ## The API equivalent If you are building this into a writing tool rather than running it in a chat, search per extracted claim and keep the design tier explicit: ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=resistance training muscle strength adults over 60" \ --data-urlencode "study_types=meta-analysis,systematic review,rct" \ --data-urlencode "human=true" \ --data-urlencode "page_size=20" ``` Each result carries `title`, `authors`, `journal_name`, `publish_year`, `doi`, `citation_count`, and `url`. On paid plans, `study_type` and `takeaway` come back too — `takeaway` is what you show the writer as the evidence quote. ## What to check before you trust it * **`NO EVIDENCE FOUND` is a real answer.** If you get none of them across a whole section, the prompt is being too accommodating — the model is stretching papers to fit. Re-run and insist on the distinction. * **Check one citation by hand per section.** Open the `url` and confirm the paper says what the table claims. Calibrate once, then trust the pattern. * **Watch for topic-level matches.** A paper about the right subject that never tested your specific claim is the most common failure. The requirement to quote a supporting sentence is what catches it. * **The tier matters for wording.** A claim supported only by observational evidence should not be written as though a trial established it. ## Related > [card] [Verify that every citation is real and supports its claim](/use-cases/citation-verification) > > The other half of the job: verify the references you already have. > [card] [Build a reusable literature-review workflow](/use-cases/literature-review-skill) > > Make this repeatable instead of pasting the prompt every time. > [card] [Best practices](/use-cases/best-practices) > > Fan-out, the filter ladder, and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Screen thousands of target-disease pairs for evidence Source: https://docs.consensus.app/use-cases/high-throughput-screening Pull literature evidence for every pair in a large matrix — the batch workload that breaks most retrieval APIs. ## What you get Evidence attached to every cell of a screening matrix. Not interactive search: a pipeline that walks tens of thousands of pairs, retrieves the literature for each, and produces a structured result you can rank and filter. ## Who it's for Computational biology and platform teams running systematic screens, and anyone whose "search" is really a batch job with a six-figure cell count. ## How Phylo runs it [Phylo](https://consensus.app/home/community-voices/phylo/) is an AI-native IDE for biology built by researchers from Stanford. One internal team uses the Consensus API to screen roughly **20,000 targets across hundreds of diseases**, pulling literature evidence for every target-disease pair. > "It's very robust, with almost zero errors. It gives you genuinely useful results, and it embeds into a high-throughput agentic workflow easily." > > — **Kexin Huang**, Co-founder & CEO, Phylo Literature is not a side feature of that product — it is underneath nearly everything it does: > "Even in analysis tasks, there's literature research underneath. For single-cell annotation, you need to find marker genes. To compare a paper's findings against your own dataset, you need literature. To brainstorm a hypothesis, to understand a mechanism of action for a particular drug and disease, all of this requires a large volume of literature. That's why we spent so much time making sure our literature search was the highest quality." > > — **Kexin Huang**, Co-founder & CEO, Phylo ## Building the screen **1. Separate the template from the entity lists** One query template crossed with two entity lists is auditable and cheap to change. Hundreds of hand-written queries are neither. **2. Decide the evidence question per cell** "Is there any literature linking this target to this disease" and "is there strong human evidence" are different screens with different filters. Pick one and hold it constant, or your cells are not comparable. **3. Pace to the rate limit, not to your ambition** This is where batch pipelines fail. The ceiling is per user, per second — see [rate limits](/use-cases/best-practices#rate-limits-and-result-caps). Build a bounded work queue with backoff and jitter before you scale the matrix, not after. **4. Cache on the full parameter set** Screens have enormous overlap — the same target appears against hundreds of diseases. Cache on every parameter with a short TTL and you cut a large fraction of the calls. **5. Score, don't just count** Paper count is a poor signal. Weight by study design, recency, and whether the evidence is human. A cell with one strong trial outranks a cell with forty in vitro papers. **6. Record the empty cells** A pair with no literature is a finding — it is either a genuine gap worth investigating or a terminology problem in your template. Distinguish them by spot-checking. ```python theme={null} def screen(targets, diseases, template, search): for target in targets: for disease in diseases: query = template.format(target=target, disease=disease) body = search(query=query, human=True, page_size=25) yield { "target": target, "disease": disease, "n": len(body["results"]), "top": body["results"][:5], # keep DOIs for provenance } ``` Pace the loop to one search per second and build in backoff, rather than firing cells in parallel. ## What to check before you trust it * **Validate the template on 50 cells first.** A template that produces noise at cell 50 produces noise at cell 50,000, expensively. * **Watch for entity-name collisions.** Short gene symbols and common disease words pull in unrelated literature. Spot-check the highest-volume cells — anomalously high counts usually mean an ambiguous name. * **Distinguish a plan cap from a real result.** If responses report finding more than they returned, your matrix is measuring your ceiling, not the literature. * **Alert when a whole run returns nothing.** Silence looks identical to a broken job. ## Related > [card] [Answer one question with many parallel searches](/use-cases/parallel-search-angles) > > The interactive counterpart to this batch pattern. > [card] [Track new papers across many topics on a schedule](/use-cases/scheduled-research-pipeline) > > Re-run a screen on a cadence and report only the delta. > [card] [Best practices](/use-cases/best-practices) > > Exhaustive sweep, rate limits, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Find human-factors evidence for a design decision Source: https://docs.consensus.app/use-cases/human-factors-evidence Pull the biomechanics, ergonomics, and usability research behind a product or workstation choice, before the decision gets defended after the fact. ## What you get The published human-subjects evidence for a design decision: what has been measured, in whom, under what conditions, with what effect size — and where the evidence simply does not exist for your population. ## Who it's for Product designers, human factors and ergonomics engineers, sports-science and footwear R\&D teams, and occupational health groups. Corporate science teams doing kinesiology and biomechanics research sit squarely here. ## Why it belongs to human-subjects rules This is the corner of applied R\&D where the evidence is about people, so the filters that matter are the clinical ones: `human`, `controlled`, and `sample_size_min` do more for result quality here than any materials or engineering filter. ## How it works **1. State the decision as an outcome question** Not "what does the research say about midsole foam" but "does increased midsole stiffness reduce metabolic cost in recreational runners". Named population, named intervention, named measurable outcome. **2. Demand human, controlled evidence** Set `human=true`, `controlled=true`, and a realistic `sample_size_min`. Biomechanics studies are often very small; a threshold around 15 to 20 filters out the least informative work without emptying the result set. **3. Match the population to your users** Trained athletes are not recreational users. Young male university students are not the general population. This literature has a well-known sampling skew, so read the population column before the finding. **4. Separate lab measures from field outcomes** A change in a gait parameter on a treadmill is not a change in injury rate in the world. Track which you have — this is where most overclaiming in this field starts. **5. Search the null result** Run the negative phrasing. In biomechanics, small studies with positive findings publish more readily than small studies without. ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=midsole longitudinal bending stiffness running economy metabolic cost" \ --data-urlencode "human=true" \ --data-urlencode "controlled=true" \ --data-urlencode "sample_size_min=15" \ --data-urlencode "domain=psych,eng,med" \ --data-urlencode "page_size=50" ``` `domain=psych,eng,med` covers most of this literature — human factors work is split across psychology, engineering, and sports medicine, and restricting to one loses a third of it. ## What to check before you trust it * **Read the sample description on every study.** Population mismatch is the dominant failure mode in this literature, and it is rarely visible in the abstract's conclusion. * **Check effect size, not significance.** A statistically significant 1% change in a lab measure may be irrelevant to a design decision. * **Watch for surrogate outcomes.** Comfort ratings, gait parameters, and injury rates are three different claims, and only the last supports an injury-reduction statement. * **Small samples dominate.** Treat any single study as provisional; look for replication before it drives a design commitment. ## Related > [card] [Substantiate a product claim with published evidence](/use-cases/claim-substantiation) > > Turn the evidence into a claim you can defend publicly. > [card] [Grade a set of papers with your own evidence rubric](/use-cases/evidence-grading-rubric) > > Score the studies consistently before deciding. > [card] [Best practices](/use-cases/best-practices) > > Filter ladder and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Build internal research tools Source: https://docs.consensus.app/use-cases/internal-tools How enterprise R&D teams, biotechs, and AI startups connect Consensus paper search to internal assistants, monitoring agents, literature tables, and scientific agents. Most organizations don't want another research app. They want the literature available *inside* what they already use — the company's ChatGPT or Claude workspace, an in-house model, a Slack bot, a spreadsheet, or the agent they ship to customers. Consensus is the retrieval layer underneath: one search call over 400+ million papers, with filters, pagination, and full-text excerpts. This guide covers the five things enterprise teams build most, how each maps to the API, and what to get right. | Build | Surface | Key parameters | Worked example | | - | - | - | - | | [Internal AI assistant](#connect-paper-search-to-your-internal-ai-assistant) | MCP, or the API as a tool | `query` plus a few filters | [Add an evidence layer](/use-cases/add-an-evidence-layer) | | [Field-monitoring agent](#keep-on-top-of-a-field) | API, scheduled | `domain`, `year_min`, `month_min` | [Track new papers on a schedule](/use-cases/scheduled-research-pipeline) | | [Exhaustive paper list](#pull-an-exhaustive-list-of-papers) | API | `page`, `page_size`, `publisher_name` | [PRISMA-ready corpus](/use-cases/prisma-corpus) | | [Literature table](#build-literature-tables-that-fill-themselves) | API, on demand or synced | `study_types`, full-text excerpts | [Drug target dossier](/use-cases/target-evidence-tiers) | | [Scientific agent](#give-your-scientific-agent-paper-search) | MCP or API | parallel queries, full-text excerpts | [Parallel search angles](/use-cases/parallel-search-angles) | ## Connect paper search to your internal AI assistant This is usually the first thing a team ships, because it takes an afternoon. People already ask the company assistant research questions. Right now it answers from memory; with Consensus attached, it answers from papers and links to them. **If your company uses ChatGPT or Claude**, add the Consensus connector with the server URL `https://mcp.consensus.app/mcp`. Each employee signs in with their own Consensus account, so their own plan and limits apply. See [Connect MCP](/consensus-mcp) for client-by-client steps. **If you run your own model or agent framework**, register `GET /v1/search` as a tool. Keep the schema small: a model chooses filters better from five well-described options than from twenty. ```python theme={null} import os, requests SEARCH_TOOL = { "name": "search_papers", "description": ( "Search 400M+ scholarly articles, including 350M peer-reviewed research papers. Use for any question whose answer " "should rest on published evidence. Returns papers with title, authors, journal, " "year, citation count, DOI, and URL. Cite only papers this tool returned." ), "input_schema": { "type": "object", "properties": { "query": {"type": "string", "description": "A natural-language research question or topic."}, "year_min": {"type": "integer", "description": "Exclude papers published before this year."}, "study_types": {"type": "array", "items": {"type": "string"}, "description": "Restrict by design, e.g. rct, meta-analysis, systematic review."}, "human": {"type": "boolean", "description": "Only include human studies."}, "sjr_max": {"type": "integer", "description": "Worst journal quartile to include. 1 keeps top-quartile journals only."}, }, "required": ["query"], }, } FIELDS = ("title", "authors", "journal_name", "publish_year", "citation_count", "doi", "url", "takeaway") def search_papers(**args): resp = requests.get( "https://api.consensus.app/v1/search", headers={"x-api-key": os.environ["CONSENSUS_API_KEY"]}, params=args, timeout=30, ) resp.raise_for_status() # Trim each result so twenty papers fit comfortably in the model's context. return [{k: p.get(k) for k in FIELDS} for p in resp.json()["results"]] ``` **What to get right** * **Tell the model when to use it and how to cite.** Put "cite only papers returned by `search_papers`, with their URL" in the system prompt, and make "the literature does not answer this" an acceptable answer. * **Return `doi` and `url` on every result** so each claim in the answer can link to its source. * **Pick the auth model deliberately.** MCP spreads load across each person's account. A shared API key concentrates every employee's searches on one rate limit — one search per second outside Enterprise. ## Keep on top of a field The request we hear most from corporate research and scouting teams: *tell me what came out this week that matters to me.* A typical example is a team tracking computer vision and robotics that wants an agent per team member, each surfacing new papers on that person's topics and following which authors and labs are publishing where. **1. Give each person or team a watchlist** A watchlist is a handful of queries plus the filters that scope them. Use `domain` to hard-filter by field (`cs`, `eng`, `bio`, `med`, and so on). **2. Window every run to the last one** Set `year_min` and `month_min` to the date of the last successful run, so each run only retrieves what is new. **3. Diff on DOI and send a digest** Store every DOI you have already sent. Each digest is the set difference, ranked by relevance to the watchlist — new papers have no citations yet, so `citation_count` is a poor signal here. **4. Track authors and institutions over time** Every result carries `authors` and `institutions`. Count them across runs to see who is publishing on your topics, where, and when an affiliation changes. ```python theme={null} import os, time, requests from collections import Counter API = "https://api.consensus.app/v1/search" HEADERS = {"x-api-key": os.environ["CONSENSUS_API_KEY"]} WATCHLISTS = { "alice": [ {"query": "vision-language-action models for robot manipulation", "domain": "cs,eng"}, {"query": "sim-to-real transfer for legged locomotion", "domain": "cs,eng"}, ], "bob": [ {"query": "event cameras for autonomous driving perception", "domain": "cs"}, ], } def digest(person, since, seen_dois): fresh = [] for watch in WATCHLISTS[person]: params = {**watch, "year_min": since.year, "month_min": since.month, "page_size": 100} resp = requests.get(API, headers=HEADERS, params=params, timeout=30) resp.raise_for_status() for paper in resp.json()["results"]: doi = paper.get("doi") if doi and doi not in seen_dois: seen_dois.add(doi) fresh.append(paper) time.sleep(1) # one search per second outside Enterprise return fresh fresh = digest("alice", since=last_run, seen_dois=load_seen("alice")) top_authors = Counter(a for p in fresh for a in p.get("authors") or []) top_labs = Counter(i for p in fresh for i in p.get("institutions") or []) ``` **What to get right** * **Keep preprints in for early signal.** Preprints are included unless you set `exclude_preprints=true`, and each result's `is_preprint` flag lets you label them in the digest. * **`journal_name` boosts, it does not filter.** It surfaces a journal higher but still returns others. To watch specific venues, pass `journal_name` and keep only matching results client-side, or use `publisher_name`, which is a hard filter. * **Log the empty runs.** A digest with nothing in it is a record worth keeping: it tells "no new papers" apart from "the job broke three weeks ago." For the scheduled version at scale, see [Track new papers across many topics on a schedule](/use-cases/scheduled-research-pipeline) and [Monitor new research about a drug, device, or brand](/use-cases/brand-and-device-monitoring). ## Pull an exhaustive list of papers Landscape reviews, due diligence, and building an internal corpus all start with the same request: *give me every paper on this.* One search returns the most relevant page; an exhaustive list means paginating to the end, scoping with hard filters, and covering the topic with more than one phrasing. ```python theme={null} def sweep(query, **filters): page, out = 0, [] while True: resp = requests.get(API, headers=HEADERS, timeout=30, params={ "query": query, "page": page, "page_size": 250, # 4 pages × 250 fills the 1,000-result window **filters, }) resp.raise_for_status() body = resp.json() out.extend(body["results"]) if body["is_end"] or body.get("next_page") is None: break page = body["next_page"] if (page + 1) * body["page_size"] > 1000: # a query returns at most 1,000 results break time.sleep(1) return out # Everything in one journal since 2020: boost with journal_name, then filter exactly. papers = sweep("perovskite solar cell stability", journal_name="Nature Energy", year_min=2020) in_journal = [p for p in papers if p.get("journal_name") == "Nature Energy"] ``` **What to get right** * **Scope with hard filters first:** `domain`, `year_min`/`year_max`, `publisher_name`, `country`, `study_types`. They shrink the set before you page through it. * **Fan out phrasings for recall.** Results are ranked by relevance to one query, so a single phrasing misses papers that use different terminology. Run several phrasings, then deduplicate on `doi`. * **Check the `page_size` echoed back.** A request above your plan's ceiling is clamped rather than rejected: 300 on Pro and Teams, 750 on Deep, 1,000 on Enterprise. * **Expect at most 1,000 results per query.** `(page + 1) × page_size` can't exceed 1,000 on any plan. For a larger landscape, split by year range, field, or sub-topic and merge on `doi`. The full method, including deduplication and flow counts, is in [Build a PRISMA-ready search corpus](/use-cases/prisma-corpus) and [best practices](/use-cases/best-practices#exhaustive-sweep). ## Build literature tables that fill themselves Drug discovery and R\&D teams often want a table rather than a chat. Rows are entities — targets, compounds, materials, indications. Columns are the literature questions you ask of every one of them. Each cell is one filtered search, an optional pass over full-text excerpts, and a short cited summary. The table fills in when someone adds a row, or refreshes on a schedule. ```python theme={null} ROWS = ["PCSK9", "ANGPTL3", "LPA"] COLUMNS = { "human_genetics": dict( template="{row} loss-of-function variants cardiovascular outcomes Mendelian randomization", human=True, ), "clinical": dict( template="{row} inhibitor randomized controlled trial outcomes", study_types=["rct", "meta-analysis"], ), "safety": dict( template="{row} inhibition adverse events safety", human=True, ), } def fill_cell(row, column): spec = dict(COLUMNS[column]) query = spec.pop("template").format(row=row) resp = requests.get(API, headers=HEADERS, timeout=30, params={ "query": query, **spec, }) resp.raise_for_status() papers = resp.json()["results"] # Your LLM call: summarize from each paper's full_text_chunks, cite by DOI. return summarize_with_citations(row, column, papers) ``` **What to get right** * **One template per column, not a hand-written query per cell.** Templates are auditable, and when you change one you know exactly which cells to refresh. * **Keep the papers behind every cell.** Store the DOIs alongside the summary so a reader can click through from any cell to its evidence. * **Use full-text excerpts for anything quantitative.** `full_text_chunks` (paid plans, open-access papers) lets a cell carry an effect size or a measured value instead of a paraphrase of the abstract. * **Refresh incrementally.** On each sync, window with `year_min` and `month_min`, and re-summarize only the cells whose paper set changed. * **Budget searches before you add rows.** Rows × columns is the number of searches per refresh: 300 targets × 5 columns is 1,500 searches, or about 25 minutes at one per second. Worked examples: [Build a drug target validation dossier](/use-cases/target-evidence-tiers), [Screen thousands of target-disease pairs](/use-cases/high-throughput-screening), and [Extract numbers from full-text papers](/use-cases/extract-values-from-full-text). ## Give your scientific agent paper search Biotech and AI startups building scientific agents use Consensus as the agent's literature tool. The agent decides what to search, Consensus returns ranked papers and excerpts, and the agent reasons over them. > [card] [Phylo — REST API, high throughput](https://consensus.app/home/community-voices/phylo/) > > Phylo's agents turn each biology question into several parallel search angles, merge the evidence, and distill it into one answer. One team screens about 20,000 targets across hundreds of diseases. > [card] [Owkin — MCP, shipped in days](https://consensus.app/home/community-voices/owkin/) > > Owkin exposed Consensus to K Pro's orchestrator the same way it exposes its other tools, replacing a curated PubMed-abstract corpus with full-text search. > "It's very robust, with almost zero errors. It gives you genuinely useful results, and it embeds into a high-throughput agentic workflow easily." > > — **Kexin Huang**, Co-founder & CEO, Phylo **Choosing a surface:** use **MCP** if your agent already consumes tools over MCP — it is a tool registration, not a new service. Use the **REST API** when you need pagination, batch workloads, or one shared quota across all your users. **What to get right** * **Search several angles per question.** One search per user question is the most common cause of thin answers. Decompose into mechanism, population, outcome, and the contradicting view, and pace the searches to your rate limit. * **Reason over excerpts, not just abstracts.** On paid plans, `full_text_chunks` gives the generation step the results paragraphs. * **Enforce grounding at generation time.** Drop any sentence that has no `doi` or `url` behind it. * **Benchmark before you switch providers.** Curate queries with known-relevant papers and measure what each provider surfaces — Phylo ranked Consensus first on its own benchmark. Worked examples: [Add a citable evidence layer to an agent](/use-cases/add-an-evidence-layer), [Answer one question with many parallel searches](/use-cases/parallel-search-angles), [Swap a hand-maintained paper corpus](/use-cases/swap-a-paper-corpus), and [Benchmark literature providers](/use-cases/benchmark-retrieval). ## Before you roll it out > [accordion] Rate limits > > The API allows one search per second on every plan except Enterprise, where limits are custom; MCP allows three per second on paid plans. Pace fan-outs and scheduled jobs, honour `retry-after` on a 429, and add jitter. See [rate limits and result caps](/use-cases/best-practices#rate-limits-and-result-caps). > [accordion] Result caps and full text > > `page_size` caps at 300 on Pro and Teams, 750 on Deep, and 1,000 on Enterprise, and a query returns at most 1,000 results across all pages. Query-relevant full-text excerpts come back on all paid plans, for the first 20 open-access papers on each page. > [accordion] Per-user or shared auth > > MCP authenticates each person with OAuth, so their own plan applies. The API uses one `x-api-key` for the whole integration — simpler to operate, but every user shares its limits. > [accordion] Caching and audit > > Cache on the full parameter set with a short TTL so new papers still land. Log every query, its filters, and the DOIs returned — it makes answers auditable and debugging far faster. > [!TIP] > > Building something your whole organization will use? An Enterprise trial gives you the higher limits and larger result pages to build without hitting ceilings. [Talk to us](https://consensus.app/home/api). ## Related > [card] [Build custom research systems](/use-cases/research-systems) > > The academic counterpart: literature reviews, writing, and reports. > [card] [Best practices](/use-cases/best-practices) > > Fan-out, the filter ladder, exhaustive sweeps, and grounding rules. > [card] [Search endpoint reference](/api-reference/search) > > Every filter and response field, with an interactive playground. > [card] [Use case library](/use-cases/gallery) > > 30 worked examples you can filter by persona. # Find the leading investigators in an indication Source: https://docs.consensus.app/use-cases/investigator-mapping Rank the authors publishing most in a therapeutic area, split by region and by recent versus lifetime output. ## What you get A ranked investigator list with paper counts, representative citations, recency, and region — built from the literature rather than from whoever your team already knows. ## Who it's for Medical affairs selecting advisory boards and speakers, clinical operations shortlisting trial sites, and business development mapping who owns a space. > [!NOTE] > > Consensus returns **author names**, not affiliations, institutions, or contact details. This produces a ranked name list you then join to a CRM, an institutional directory, or a trial registry. Treat it as the discovery step, not the whole workflow. ## How it works **1. Sweep the indication widely** Fan out across the indication, its mechanisms, its endpoints, and its main comparators. Paginate — emerging investigators sit in the long tail, and the top 20 results only ever show you the established names. **2. Count authors across the deduplicated set** Tally every name, weighting by `citation_count`, recency, and `sjr_best_quartile`. Track first and last author positions separately where you can infer them: last author usually signals the lab, first author the person doing the work. **3. Rank twice** Once by lifetime output, once by output in the last 24 months. The two lists differ substantially, and the second is usually the more actionable — it tells you who is active now rather than who was. **4. Add geography** Use `country` to build per-region shortlists. A single global list is rarely what a regional medical team can act on. **5. Read one paper per name** Output volume is not fit. Check that their actual work matches the specific question before anyone gets contacted. ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=CAR-T therapy relapsed refractory multiple myeloma outcomes" \ --data-urlencode "human=true" \ --data-urlencode "year_min=2023" \ --data-urlencode "country=us,gb,de" \ --data-urlencode "page_size=200" ``` ```python theme={null} from collections import Counter authors = Counter() for paper in results: weight = 1 + (paper.get("citation_count", 0) ** 0.5) / 10 for name in paper.get("authors", []): authors[name] += weight ``` ## What to check before you trust it * **Name ambiguity is the main failure.** Common surnames merge distinct researchers, and the same person appears under several spellings. Verify anyone you plan to act on. * **Volume rewards large consortia.** Authors on big multi-site trials accumulate counts without being the intellectual lead. Weight last-author position where you can. * **Check the recency split before using the lifetime list.** Highly-cited historical names may have left the field entirely. * **This is a discovery list, not a vetted one.** Conflicts of interest, availability, and current affiliation all need checking elsewhere. ## Related > [card] [Monitor new research about a drug, device, or brand](/use-cases/brand-and-device-monitoring) > > Keep watching the space and the names in it. > [card] [Build a drug target validation dossier](/use-cases/target-evidence-tiers) > > Go deep on the science these investigators are publishing. > [card] [Best practices](/use-cases/best-practices) > > Fan-out and merge, and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Build and maintain a Library collection from your assistant Source: https://docs.consensus.app/use-cases/library-reading-list Have the Consensus research agent find papers, save them into a Library collection, then audit, filter, and rename that collection over MCP or the API. ## What you get A curated [Library](/core-features/library) collection you never had to click together. The research agent finds the papers, saves them into the collection you name, and later lists, filters, and summarizes what's in it. Every change shows up in Consensus right away, so you can keep reading and sharing in the web app. ## Who it's for Researchers who live in Claude, ChatGPT, or another MCP client and want their reading list to land in Consensus. It needs a personal Consensus account or API key: organization API keys have no Library. ## Step 1: Find papers and save them Ask for the research and the save in one brief. Name a new collection explicitly. The agent never invents one. ```text theme={null} Use Consensus to find 10 randomized controlled trials from the last 10 years on magnesium supplementation and sleep quality in adults. Summarize what they found, then save all 10 to a new Library collection called "Magnesium and Sleep". ``` The answer says how many papers it saved and how many were already in the collection. ## Step 2: Read the collection back Later, ask what's in it. The agent can list your collections, list a collection's papers filtered by publication year, date added, or a title, author, or journal match, and search within it. ```text theme={null} Use Consensus to list my "Magnesium and Sleep" collection. Which papers were published before 2020, and which did I add in the last month? ``` ```text theme={null} Use Consensus to list all of my Library collections and how many papers each has. ``` ## Step 3: Update it Extend or rename the collection in a follow-up on the same Thread, or in a new one. ```text theme={null} Find any meta-analyses on magnesium and sleep published since 2022 that aren't in my "Magnesium and Sleep" collection, and add them. Then rename the collection to "Magnesium and Sleep - review". ``` ## The API equivalent Send the same brief to `POST /v1/threads` with a personal API key, then poll the Thread for the answer: ```bash theme={null} curl -X POST "https://api.consensus.app/v1/threads" \ -H "x-api-key: $CONSENSUS_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "input_message": "Find 10 randomized controlled trials from the last 10 years on magnesium supplementation and sleep quality in adults. Summarize what they found, then save all 10 to a new Library collection called Magnesium and Sleep." }' ``` To focus a later Thread on the collection, pass its id as `attachments.collection_ids`. The id is the number in the collection's web URL, `consensus.app/library/{id}/`: ```json theme={null} { "input_message": "Across these trials, which doses and durations improved sleep onset latency?", "attachments": { "collection_ids": [123] } } ``` See [Work in your Library](/api-quickstart-threads#work-in-your-library) for the full list of actions. ## What to check before you trust it * **Saves don't ask first.** The agent saves as soon as the brief says to. Say exactly which papers and which collection you mean. * **Names must match.** A collection name the agent can't find comes back as not found. Ask it to list your collections if you're unsure of the exact name. * **Repeated names are ambiguous.** If two collections share a name, the agent picks one for saves and listings. Name the parent collection too. * **Removing, deleting, and moving stay in the web app.** They need a confirmation step, which MCP and the API don't have. * **Use a personal key.** Threads started with an organization API key have no Library. ## Related > [card] [Turn a syllabus into a recommended reading list](/use-cases/syllabus-reading-list) > > Build the reading list first, then save it to a collection. > [card] [create_thread](/mcp-tools/create-thread#library-actions) > > Every Library action the research agent supports over MCP. # Build a reusable literature-review workflow Source: https://docs.consensus.app/use-cases/literature-review-skill A repeatable review workflow that picks a search framework, allocates a search budget, and synthesizes findings into a research guide your whole team can run the same way. ## What you get A **launching pad**, not a finished review: a document that orients a researcher in an unfamiliar field fast enough to start reading confidently. What a generous colleague who knows the field would tell you over coffee — here's the lay of the land, here are the key people, here's how the thinking evolved, read these first. Consensus ships this as a packaged skill. The workflow below is what it does, so you can run it, adapt it, or rebuild it in Claude Code, Codex, or your own agent framework. ## Who it's for Research-software engineers, methodologists, and technical researchers who run the same class of review repeatedly. Teams doing this report cutting a review from roughly two days to one. ## Set up the connection ```bash Claude Code theme={null} claude mcp add --transport http consensus https://mcp.consensus.app/mcp ``` ```bash Codex theme={null} codex mcp add consensus --url https://mcp.consensus.app/mcp codex mcp login consensus ``` ## Phase 1 — Reconnaissance Run **one** broad exploratory search. Confirm it returned data before going further. Then read the abstracts for: * The major themes and subfields * **The terminology researchers actually use** — usually not the phrasing the requester used * Methodological distinctions that matter here (RCT vs observational, animal vs human) * Angles the requester probably has not considered Note citation counts as they come back. A paper with unusually high citations relative to its age is likely foundational — flag it for the priority reading list. ## Phase 2 — Pick a framework Evaluate **PICO first**; it applies far more broadly than just clinical questions. | Framework | Use when | Components | | - | - | - | | **PICO** (primary) | Health, clinical, behavioural, educational, and most social science questions | Population · Intervention · Comparison · Outcome | | **SPIDER** | No clear intervention or comparison group — lived experience, attitudes, perceptions | Sample · Phenomenon of Interest · Design · Evaluation · Research type | | **Decomposition** | The topic is a technology or system rather than a population or behaviour | Core mechanism · Applications · Limitations · Alternatives | Many real questions span frameworks. "How does social media affect teen mental health" has PICO elements, qualitative elements, and technology elements. Pick a primary for structure and say which components borrow from elsewhere — the goal is clarity, not orthodoxy. Name the chosen framework explicitly, show how the topic maps to each component, and say in one sentence why you picked it over the alternatives. ## Phase 3 — Confirm before spending the budget Stop and check in. Output a scannable summary: 1. **What the literature shows** — 3–4 sentences on themes, terminology, what is well-studied, what is contested. 2. **Framework mapping table** — one row per component, each proposing a sub-area to explore, plus a fifth row for any cross-cutting theme. 3. **Search depth** — let the requester choose, and say what constrains it. | Depth | Searches | Allocation | | - | - | - | | Quick scan | 5 | 5 sub-area searches; skip era-gated and review-specific passes | | **Standard** (recommended) | 10 | 5 sub-area · 2 review-article · 2 era-gated · 1 follow-up on the highest-cited paper | | Deep dive | 20 | 5 sub-area · 5 review-article · 4 era-gated · 3 follow-ups on top-cited papers · 3 spare for emerging threads | The key idea: extra budget buys **deeper analysis**, not more of the same. Review-article searches (`systematic review [topic]`, `meta-analysis [topic]`) are disproportionately valuable because one good review pre-digests dozens of primary studies. ## Phase 4 — Execute Run searches **sequentially** — send, wait for the result, confirm it contains data, record what came back, then send the next. There is a one-query-per-second limit on every plan except Enterprise, so parallel firing produces 429s rather than speed. See [rate limits](/use-cases/best-practices#rate-limits-and-result-caps). **Era-gated searches** are the cheapest way to understand a field's trajectory: run the most important sub-area once with `year_max: 2015` and once with `year_min: 2021`, then compare. * **Terminology shifts** — "gut flora" became "gut microbiome". A researcher searching only modern terms will miss foundational older work. * **Conclusion shifts** — older papers reaching different conclusions signals a paradigm shift or accumulating evidence. * **Methodological evolution** — observational studies → RCTs → meta-analyses tells you how mature the evidence base is. ### Cross-search intelligence This is what turns a pile of results into field knowledge. Track three things across **all** searches: 1. **Repeat-hit papers.** A paper appearing in 3 of 5 sub-area searches is foundational to the whole field, not just one sub-area. Flag these as must-reads. 2. **Recurring authors.** The same author group across multiple searches signals a dominant lab. Note the top 3–5 — anyone entering this field needs to know the key voices. 3. **Citations per year.** Divide citation count by years since publication. A 2023 paper with 150 citations is a far stronger signal than a 2008 paper with 150. ## Phase 5 — Produce the guide Structure the output as a launch pad: 1. **Topic overview** — one tight paragraph: what the topic is, which framework was used, and a characterization of the evidence landscape ("robust on X, sparse on Y"). 2. **Start here — priority reading order.** The most actionable section. 5–7 papers across all sub-areas, ordered as a newcomer should read them: best recent review first, then the foundational paper(s), then 2–3 at the current frontier, ending with one that highlights a gap or controversy. Per paper give what it contributes *in the sequence* and what to pay attention to while reading it ("focus on Table 3, which compares effect sizes across all RCTs"). 3. **How the field got here** — a 5–8 row timeline plus a note on terminology evolution. 4. **Sub-area guides** — per sub-area: what the research shows, 3–5 key papers, 6–10 search terms including historical ones, and 2–3 ready-to-paste Boolean strings. 5. **Key research groups** — the 3–5 most frequent authors, their sub-areas, a representative paper each. 6. **Open questions and gaps** — split into methodological, population/context, and conceptual. For each, say *why it matters*, not just that it is unstudied. 7. **Bibliography** — alphabetical, every entry with a clickable full Consensus URL. 8. **Audit log.** ## Audit requirements Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability). For this workflow, log per search: query, filters as sent, papers returned, status. Then report searches executed, searches successful, searches failed after retry, unique papers received after dedup, and papers cited. State the detected plan cap and what it means for coverage: at 10 successful searches with a 10-result cap, the theoretical ceiling is 100 papers before dedup. Flag any sub-area returning fewer than 5 papers as potentially incomplete coverage rather than a genuine gap. ## What to check before you trust it * **Confirm the reading order is actually ordered.** If it is just five good papers with no sequencing logic, the most valuable section has degraded into a list. * **Check the terminology note.** A field whose vocabulary shifted and whose guide does not mention it will send researchers into incomplete searches. * **Watch for silent filter emptiness.** If a narrow pass returned almost nothing, relax one filter at a time and say which was binding — never present that as "little evidence exists". * **Calibrate once.** Have a person check one sub-area's papers by hand before trusting the method at volume. ## Related > [card] [Turn a syllabus into a recommended reading list](/use-cases/syllabus-reading-list) > > The teaching-focused sibling of this workflow. > [card] [Track new papers across many topics on a schedule](/use-cases/scheduled-research-pipeline) > > Take the same method and run it unattended on a cron. > [card] [Best practices](/use-cases/best-practices) > > Rate limits, grounding rules, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Compare candidate materials against published test data Source: https://docs.consensus.app/use-cases/materials-comparison Search published performance data for each candidate under your actual operating conditions, and find out which combinations nobody has tested. ## What you get A comparison matrix of candidates against the conditions you care about. Every populated cell cites a paper and the conditions the value was measured under. Every empty cell is a gap — and those are where your test budget should go. ## Who it's for Materials, process, and reliability engineers choosing between options with incomplete internal data. Semiconductor, automotive, aerospace, and consumer-product R\&D groups. ## The move that makes this work Express the requirement as **conditions**, not as a material name. The spec becomes the query: temperature range, cycle count, atmosphere, dimension, contaminant, load. Then hold those conditions constant across every candidate so the results are comparable. A search for a material name returns everything ever written about it. A search for the material under your conditions returns the handful of papers that answer your question. ## How it works **1. Turn the spec into physical conditions** Write the operating envelope out explicitly before searching. This is the step that determines whether the output is usable. **2. One fan-out per candidate, conditions held constant** Run a query set per material or process, varying only the candidate. Anything else and you are comparing searches, not materials. **3. Set the domain, or drown** Without `domain=mat,eng,phys,chem`, technical vocabulary pulls in unrelated biomedical literature. This single parameter is the difference between a usable and an unusable result set. **4. Extract values with their conditions attached** Read the excerpts in `full_text_chunks`, which paid plans return. A performance number without its test conditions is not comparable to another number, and conditions live in methods sections rather than abstracts. **5. Report the gaps as loudly as the data** Say plainly which candidate-condition combinations have no published data. That is the most actionable output of the whole exercise. ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=copper hybrid bonding void formation thermal cycling reliability" \ --data-urlencode "domain=mat,eng,phys" \ --data-urlencode "page_size=50" ``` > [!NOTE] > > Full-text excerpts come back on paid plans and cover open-access papers only, up to the first 20 on each page. Engineering and materials literature skews toward closed-access conference proceedings, so expect coverage here to be patchier than in biomedicine. Check before treating the matrix as complete. Useful `domain` codes for this work: `mat` (materials), `eng` (engineering), `phys` (physics), `chem` (chemistry), `env` (environmental). ## What to check before you trust it * **Verify conditions on every extracted value.** A modulus measured at room temperature says nothing about behaviour at 200°C, and the two look identical in a table. * **Watch for standards mismatch.** Values measured under different test standards are not comparable even when units match. * **Empty cells may be a terminology problem.** Before reporting a gap, re-run with the alternative names the field uses for the same material or process. * **Published performance is best-case.** Laboratory conditions and production conditions differ; treat literature values as an upper bound to validate, not a spec to adopt. ## Related > [card] [Find published process conditions for a manufacturing step](/use-cases/process-window-precedent) > > Once you have chosen the material, find the conditions to run it at. > [card] [Extract numbers and formulas from full-text papers](/use-cases/extract-values-from-full-text) > > The extraction discipline this depends on. > [card] [Best practices](/use-cases/best-practices) > > Excerpt-first extraction and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Answer one question with many parallel searches Source: https://docs.consensus.app/use-cases/parallel-search-angles Decompose a research question into multiple search angles, gather evidence across all of them, and distill one cited answer. ## What you get The retrieval pattern behind a good "ask a research question" feature: one user question becomes many angles, each angle its own search, and the results merge into a single summary the user can trace back to papers. ## Who it's for Product engineers building research answers into an application — the difference between this and a single search is the difference between a demo and a product. ## How Phylo runs it [Phylo](https://consensus.app/home/community-voices/phylo/)'s agents make parallel calls across multiple angles of a question, gather the relevant papers, and distill the results into a summary for the user. The workflow starts wherever the researcher is: | Step | What happens | | - | - | | 1 | A biology question is defined — a hypothesis, a GWAS hit, or a dataset signal | | 2 | The agent generates multiple search angles for better coverage | | 3 | Consensus returns evidence across those retrieval paths | | 4 | The agent distills it into a usable research answer | > "It's very robust, with almost zero errors. It gives you genuinely useful results, and it embeds into a high-throughput agentic workflow easily." > > — **Kexin Huang**, Co-founder & CEO, Phylo ## Generating good angles This is where the quality lives. A weak decomposition produces five rephrasings of the same query; a strong one covers genuinely different retrieval paths. **1. Vary the axis, not just the wording** Cover the mechanism, the population, the outcome, the competing explanation, and the methodology. Five paraphrases of one angle retrieve nearly the same papers and give false confidence in coverage. **2. Include the null and the contradiction** Add an angle phrased toward the opposite finding. Searching only for confirmation returns confirmation, and users notice when your product never disagrees with them. **3. Use the field's terminology, not the user's** The user's phrasing is rarely what the literature indexes under. Translate first — a reconnaissance search to learn the vocabulary, then the real angles. **4. Bound the concurrency to your tier** "Parallel" here means angles, not simultaneous calls: there is a one-query-per-second limit on every plan except Enterprise, so issue the searches sequentially. See [rate limits](/use-cases/best-practices#rate-limits-and-result-caps). **5. Merge on DOI and keep the angle that found it** A paper surfaced by several independent angles is a stronger signal than one found by a single query — that count is worth surfacing in the answer. ## Distilling without losing the thread The synthesis step is where grounding usually breaks: * **Group by finding, not by paper.** A paper-by-paper summary is a reading list; grouping by finding with the supporting papers attached is an answer. * **Attach a citation to every claim** and drop sentences you cannot support. Make "the literature does not settle this" a real output. * **Show which angles were searched.** It tells the user what was and was not covered, and it is the cheapest trust-building thing you can render. * **Report disagreement.** Where angles returned conflicting evidence, say so rather than averaging it into a confident middle. ## What to check before you trust it * **Inspect the generated angles on real questions.** If they are paraphrases, fix the decomposition prompt — no amount of retrieval quality compensates. * **Test a question with a genuinely contested answer.** The output should surface the conflict, not resolve it silently. * **Watch merge behaviour on near-duplicates.** Preprint and published versions of the same work can carry different DOIs. * **Measure it.** Use a [benchmark](/use-cases/benchmark-retrieval) to check whether more angles actually improve recall, or just cost more. ## Related > [card] [Screen thousands of target-disease pairs for evidence](/use-cases/high-throughput-screening) > > The batch counterpart, on the same infrastructure. > [card] [Add a citable evidence layer to an agent you already ship](/use-cases/add-an-evidence-layer) > > Wiring this into an agent you have already built. > [card] [Best practices](/use-cases/best-practices) > > Fan-out and merge, rate limits, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Assemble a payer evidence dossier Source: https://docs.consensus.app/use-cases/payer-evidence-dossier Pull the cost-effectiveness, burden-of-illness, and real-world evidence a reimbursement submission needs, in one pass. ## What you get A structured evidence pack spanning four literatures that rarely sit together: clinical efficacy, burden of illness, economic evaluation, and real-world outcomes — each section cited, with the gaps named. ## Who it's for HEOR and market access teams building reimbursement submissions, and the consultancies who build them on their behalf. ## Why this one is awkward Payer dossiers span clinical and economic literature that use different vocabularies, sit in different journals, and are indexed under different fields. A single query returns one of them and misses the rest. The fix is running four deliberately separate searches and keeping them separate in the output. ## The four searches | Section | What you are looking for | Filters that help | | - | - | - | | **Clinical efficacy** | Trials and meta-analyses establishing the effect | `study_types=rct,meta-analysis`, `human=true` | | **Burden of illness** | Prevalence, mortality, quality of life, productivity loss | `domain=med,econ`, `study_types=cohort study,cross-sectional study` | | **Economic evaluation** | Cost-effectiveness, cost-utility, budget impact | `domain=econ,med` | | **Real-world evidence** | Registry, claims, and observational outcomes outside trials | `study_types=cohort study,longitudinal / panel data study` | ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=cost-effectiveness analysis quality-adjusted life year heart failure treatment" \ --data-urlencode "domain=econ,med" \ --data-urlencode "human=true" \ --data-urlencode "page_size=50" ``` Add `country` when the submission is to a specific national payer — cost-effectiveness evidence is jurisdiction-dependent, and a US cost-per-QALY carries little weight with a European HTA body. ## The prompt ```text theme={null} You are assembling the evidence base for a reimbursement dossier. Report each section separately; do not merge them into a single narrative. Product: [NAME] Indication: [INDICATION] Payer / jurisdiction: [MARKET] 1. Clinical efficacy — the effect, its size, its certainty, in whom. 2. Burden of illness — prevalence, mortality, quality of life, and economic burden in the target jurisdiction where available. If the evidence is from another country, say which and flag the transfer as an assumption. 3. Economic evaluation — published cost-effectiveness work, with the model type, perspective, time horizon, and threshold used. Note when a study's jurisdiction, currency, or year makes it non-comparable. 4. Real-world evidence — outcomes outside trial conditions, and how they differ from trial results. Then: - Where the four sections disagree, and the most likely reason. - Evidence gaps a payer is likely to challenge, stated plainly. - What additional evidence would most strengthen the submission. Cite everything with journal, year, and DOI. ``` ## Audit requirements Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability). Log the four search sets separately with their counts. If a section returned little — economic evaluations are frequently sparse for newer products — that sparsity is a finding a payer will also notice, so report it rather than padding. ## What to check before you trust it > [!WARNING] > > A literature pack is an input to a submission, not a submission. HTA requirements are jurisdiction-specific and prescriptive; the dossier's structure and evidence standards come from the payer's own guidance, not from this workflow. * **Check jurisdiction on every economic study.** Cost-effectiveness does not transfer across health systems, and a reviewer will spot a mismatched threshold immediately. * **Check currency and year.** Costs from a decade ago in another currency need conversion and inflation adjustment before they mean anything. * **Read the gap section first.** Where the evidence is thin is exactly where the payer will push, so it is the most useful output. * **Real-world and trial evidence disagreeing is normal.** Explain the gap rather than presenting only the flattering one. ## Related > [card] [Monitor new research about a drug, device, or brand](/use-cases/brand-and-device-monitoring) > > Keep watching the evidence base after submission. > [card] [Grade a set of papers with your own evidence rubric](/use-cases/evidence-grading-rubric) > > Score the evidence you assemble consistently. > [card] [Best practices](/use-cases/best-practices) > > Fan-out and merge, and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Build a PRISMA-ready search corpus Source: https://docs.consensus.app/use-cases/prisma-corpus Assemble a reproducible, deduplicated record set with the flow counts a systematic review has to report. ## What you get A deduplicated corpus, a search strategy locked before results were seen, and the counts PRISMA asks for: records identified, duplicates removed, records screened, records excluded with reasons. > [!NOTE] > > This is a recipe built on the API, not a packaged product. Consensus supplies the search, filters, and pagination; you own the strategy log and the flow diagram. It is one of the most requested workflows we do not yet ship end to end. ## Who it's for Methodologists and research-software teams producing a corpus that has to survive peer review, plus anyone publishing where PRISMA reporting is mandatory. ## Lock the strategy first Reproducibility is decided before you search, not after. Write down and persist: * the full query list * every filter, exactly as it will be sent * the inclusion and exclusion criteria * the date the search was run A strategy adjusted after seeing results is a search, not a method. If you must change it, version it and report both. ## Sweep to exhaustion ```python theme={null} import os, requests API = "https://api.consensus.app/v1/search" HEADERS = {"x-api-key": os.environ["CONSENSUS_API_KEY"]} STRATEGY = dict( study_types=["rct", "meta-analysis", "systematic review"], human=True, year_min=2015, exclude_preprints=True, page_size=200, ) def sweep(query): page, out = 0, [] while True: r = requests.get(API, headers=HEADERS, params={"query": query, "page": page, **STRATEGY}) r.raise_for_status() body = r.json() out.extend(body["results"]) if body["is_end"] or body.get("next_page") is None: break page = body["next_page"] if (page + 1) * body["page_size"] > 1000: # a query returns at most 1,000 results break return out corpus, per_query = {}, {} for query in QUERY_SET: # your locked list found = sweep(query) per_query[query] = len(found) # pre-dedup count, needed for PRISMA for paper in found: corpus.setdefault(paper["doi"], paper) ``` > [!NOTE] > > Pagination and page sizes above 20 need a paid plan or Enterprise key; `page_size` caps at 300 (Pro, Teams), 750 (Deep), higher on Enterprise. Over-large requests are clamped, not rejected — read the `page_size` echoed back. See [rate limits](/use-cases/best-practices#rate-limits-and-result-caps). ## Deduplicate and count Merge on `doi`, keeping the first-seen record plus which queries found it. Retain **both** the pre- and post-deduplication totals — PRISMA reports each separately, and you cannot reconstruct them later. Records without a DOI need a fallback key (normalized title plus year). Count how many took that path; a large number signals a coverage or record-quality problem worth mentioning. ## What PRISMA wants back | Flow stage | Where it comes from | | - | - | | Records identified | Sum of per-query pre-dedup counts | | Duplicates removed | Pre-dedup total minus unique DOIs | | Records screened | Size of the deduplicated corpus | | Records excluded, with reasons | From [abstract screening](/use-cases/abstract-screening), per criterion | | Full-text assessed | Records escalated past abstract screening | | Studies included | Final set | > [!WARNING] > > Consensus is one source. A review claiming comprehensive coverage will normally be expected to search additional databases and registries, and reviewers will ask which ones. Report Consensus as one search source with its date and strategy, not as the whole search. ## Audit requirements Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability). Persist per query: the query text, filters exactly as sent, pages walked, records returned, and the run date. That table *is* your methods section's evidence. ## What to check before you trust it * **Confirm you actually reached the end.** A query returns at most 1,000 results in total. A sweep that stopped at that window with `is_end` false has truncated results — say so, or split the query and re-run. * **Check the dedup key.** Preprint and published versions of the same study carry different DOIs and will both survive. * **Re-run the strategy on the reporting date.** Literature moves; reviewers expect a search date, and a stale corpus invites a "please update" revision. ## Related > [card] [Screen abstracts against inclusion criteria](/use-cases/abstract-screening) > > The next step, with the exclusion reasons PRISMA needs. > [card] [Write the search-strategy section of your methods](/use-cases/search-strategy-methods) > > Turn the strategy log into publishable prose. > [card] [Best practices](/use-cases/best-practices) > > Exhaustive sweep, rate limits, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Find published process conditions for a manufacturing step Source: https://docs.consensus.app/use-cases/process-window-precedent See what temperature, time, atmosphere, and concentration ranges others have published for a process, so you start from precedent rather than from scratch. ## What you get A table of published process conditions for your step — each with the material system it applied to, the outcome achieved, and the paper it came from. A starting envelope for your own DOE instead of a blank one. ## Who it's for Process engineers standing up a new process, transferring one between sites or tools, or trying to widen a window that keeps drifting out of spec. ## Why this needs full text Process conditions live in methods sections. Abstracts report that a process worked; they almost never report the temperature ramp, the dwell, the atmosphere, or the concentration. This use case depends on the full-text excerpts in `full_text_chunks` and on phrasing the query as the conditions themselves. > [!NOTE] > > Full-text excerpts come back on paid plans and cover open-access papers only, up to the first 20 on each page. Process literature is heavily represented in closed conference proceedings, so expect partial coverage and treat what you get as a starting point rather than a survey. ## The extraction schema Fix these columns before you start, or you will end up with values you cannot compare: | Column | Why | | - | - | | Process step | The specific operation, not the overall flow | | Parameter and value | With units exactly as published | | Range or single point | A window is far more useful than one setting | | Material system | Conditions rarely transfer across substrates | | Tool or method | Equipment differences change achievable ranges | | Outcome achieved | The condition means nothing without what it produced | | Source | DOI plus the excerpt the value came from | ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=atomic layer deposition alumina process temperature precursor pulse time conditions" \ --data-urlencode "domain=mat,eng,chem" \ --data-urlencode "page_size=100" ``` Phrase the query as the parameter you want, not the process in general — excerpts are selected for relevance to your query, so `deposition temperature precursor pulse` returns methods paragraphs while the bare process name returns introductions. ## What to check before you trust it * **Never transfer conditions without checking the material system and tool.** Published conditions are specific to a substrate, a reactor geometry, and a precursor. They are a starting point for a DOE, not a recipe. * **Watch for missing parameters.** Papers routinely omit one condition that turns out to be load-bearing. Mark those as NOT REPORTED rather than assuming a standard value. * **Prefer papers reporting a window over a single point.** A range tells you the sensitivity; a point tells you almost nothing about robustness. * **Check the outcome definition.** "Successful deposition" means different things across groups; align on the metric before comparing conditions. ## Related > [card] [Extract numbers and formulas from full-text papers](/use-cases/extract-values-from-full-text) > > The general extraction discipline underneath this. > [card] [Find published work on a failure or degradation mechanism](/use-cases/failure-mode-scan) > > When the process window keeps producing defects. > [card] [Best practices](/use-cases/best-practices) > > Excerpt-first extraction and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Build custom research systems Source: https://docs.consensus.app/use-cases/research-systems How academics and research groups use the Consensus API and MCP to build literature-review, writing, and reporting workflows they can reproduce. Academic teams build on Consensus for a different reason than companies do. The goal isn't to put the literature inside a product — it's to make your own research faster and more reproducible. That usually takes one of two shapes: * **In code** — a Python script, a notebook, or a Claude Code or Codex project that runs many searches, keeps a log, and writes the output. * **In tools you already use** — Consensus connected over MCP to Claude or ChatGPT, driven by a reusable prompt or skill. Both use the same search, and both follow the same rules that make the output something you can defend. | Build | Typical form | Worked examples | | - | - | - | | [Literature-review pipeline](#literature-review-pipelines) | Script or skill | [Review workflow](/use-cases/literature-review-skill), [PRISMA corpus](/use-cases/prisma-corpus) | | [Citation-grounded writing](#writing-with-grounded-citations) | MCP or script | [Ground a manuscript](/use-cases/ground-a-manuscript), [Verify citations](/use-cases/citation-verification) | | [Evidence reports](#evidence-reports-on-demand-or-on-a-schedule) | Script, scheduled or on demand | [Scheduled pipeline](/use-cases/scheduled-research-pipeline), [Biomarker report](/use-cases/biomarker-report) | | [No-code research](#use-the-tools-you-already-have) | MCP in Claude or ChatGPT | [Connect MCP](/consensus-mcp) | ## Literature-review pipelines A good review is many searches, not one. Decompose the question into angles, run each with the same filters, merge on DOI, and keep a log of everything you did. The log becomes your methods section. ```python theme={null} import os, time, requests from datetime import date API = "https://api.consensus.app/v1/search" HEADERS = {"x-api-key": os.environ["CONSENSUS_API_KEY"]} ANGLES = [ "resistance training cognitive function older adults", "strength training executive function aging randomized trial", "exercise intervention memory elderly meta-analysis", "resistance exercise mild cognitive impairment", "resistance training no effect on cognition in older adults", # the null angle ] FILTERS = dict(human=True, study_types=["rct", "meta-analysis", "systematic review"], year_min=2010) corpus, log = {}, [] for query in ANGLES: resp = requests.get(API, headers=HEADERS, params={"query": query, **FILTERS}, timeout=30) resp.raise_for_status() results = resp.json()["results"] log.append({"date": date.today().isoformat(), "query": query, "filters": FILTERS, "returned": len(results)}) for paper in results: key = paper.get("doi") or paper["url"] corpus.setdefault(key, {**paper, "found_by": []})["found_by"].append(query) time.sleep(1) # one search per second outside Enterprise print(f"{len(ANGLES)} queries, {sum(e['returned'] for e in log)} results, {len(corpus)} unique papers") ``` **What to get right** * **Vary the angle, not just the wording.** Cover mechanism, population, outcome, and the contradicting finding. Five paraphrases of one idea retrieve the same papers and give false confidence in coverage. * **Tighten filters one step at a time.** Start broad, then add study design, population, and journal quality, recording the count at each step. See the [filter ladder](/use-cases/best-practices#filter-ladder). * **Count queries, results, and cited papers separately.** They are three different numbers, and reviewers will ask for all of them. * **Keep `found_by`.** How many angles surfaced a paper is a useful relevance signal, and it documents your search. Go deeper: [Build a reusable literature-review workflow](/use-cases/literature-review-skill), [Build a PRISMA-ready search corpus](/use-cases/prisma-corpus), [Screen abstracts against inclusion criteria](/use-cases/abstract-screening), and [Write the search-strategy section of your methods](/use-cases/search-strategy-methods). ## Writing with grounded citations The second most common system: match every claim in a draft to a real paper, or flag it as unsupported. It works in either direction — finding citations for a draft, or checking that an existing reference list says what the text claims. **1. Split the draft into claims** One sentence-level claim per row. Claims that bundle two findings need two rows. **2. Search each claim as a query** Phrase the claim itself as the query, with filters that match the strength it asserts. A claim about an effect in humans deserves `human=true`. **3. Judge support, not topic** Have the model check each candidate's `takeaway`, abstract, or full-text excerpts against the claim. A paper on the right topic that doesn't support the claim is the error that survives review. **4. Output a table** Claim, supporting paper with link, and a verdict: supported, partially supported, or unsupported. Unsupported is a useful result — it tells you what to soften or cut. Go deeper: [Find a real citation for every claim in a draft](/use-cases/ground-a-manuscript), [Verify that every citation is real and supports its claim](/use-cases/citation-verification), and [Find which journals to submit to](/use-cases/where-to-submit). ## Evidence reports on demand or on a schedule Once a search strategy works, run it again. Labs use this for a weekly new-papers digest on their topic, a standing evidence report on a biomarker or intervention, or a reading list regenerated each term. * **On demand:** the same script, parameterized by topic, run whenever someone needs the report. * **On a schedule:** a cron job or GitHub Action that windows each run with `year_min` and `month_min`, diffs on DOI against the last run, and reports only what is new. Go deeper: [Track new papers across many topics on a schedule](/use-cases/scheduled-research-pipeline), [Create a biomarker evidence report](/use-cases/biomarker-report), [Turn a syllabus into a reading list](/use-cases/syllabus-reading-list), and [Find NIH grants for a research idea](/use-cases/grant-finder). ## Use the tools you already have You don't need to write code. [Connect Consensus over MCP](/consensus-mcp) to Claude, ChatGPT, or Claude Code, and the assistant runs the searches for you. The quality comes from the prompt, so write down the method once and reuse it: ```text theme={null} Search Consensus for randomized trials and meta-analyses since 2015 on resistance training and cognition in adults over 60. Run at least five searches from different angles, including one looking for null results. Deduplicate the results, then give me: 1. An evidence table: study, design, sample size, main finding, link. 2. A log of every search you ran, with its filters and how many results it returned. Cite only papers returned in this session. ``` To share a method across a lab, package the prompt as a skill so everyone runs it the same way. See [Build a reusable literature-review workflow](/use-cases/literature-review-skill). ## What makes it rigorous > [accordion] Cite only what came back in this session > > Never supplement results with the model's memory or a paper you happen to know. Every citation needs a DOI or URL returned by a search you ran. > [accordion] Log enough to reproduce it > > Record each query, its filters, the date, and the result count. That is what turns a search into a methods paragraph, and what lets someone else rerun it. > [accordion] Know your plan's result cap > > Over the API, Free returns up to 20 results per request, Pro and Teams up to 300, and Deep up to 750; over MCP, Free returns 10. A thin result set can be a plan ceiling rather than a gap in the literature — say which it was. > [accordion] Pace your searches > > The API allows one search per second on every plan except Enterprise, and MCP three per second on paid plans. Run fan-outs sequentially and retry once on failure; never silently skip a failed search. The full set of rules, including the audit-log format, is in [best practices](/use-cases/best-practices#grounding-and-auditability). ## Related > [card] [Build internal research tools](/use-cases/internal-tools) > > The organizational counterpart: assistants, monitoring agents, and agents. > [card] [Best practices](/use-cases/best-practices) > > Fan-out, the filter ladder, exhaustive sweeps, and grounding rules. > [card] [Get started with the API](/api-get-started) > > Your first authenticated request, in seven languages. > [card] [Use case library](/use-cases/gallery) > > 30 worked examples you can filter by persona. # Track new papers across many topics on a schedule Source: https://docs.consensus.app/use-cases/scheduled-research-pipeline Sweep hundreds of topics on a cron, diff against the last run, and report only what changed. ## What you get An unattended job that re-runs a fixed query set on a schedule, stores what it has seen, and produces a delta — the papers that are new since last time, ranked by whether they change a decision you have already made. One research team runs this across 121 countries every six hours to track education-reform literature, feeding a public site. The shape generalizes to any entity list: compounds, competitors, materials, indications, regions. ## Who it's for Engineers and technical researchers maintaining a living evidence base rather than answering a one-off question. Runs from Codex, Claude Code, a GitHub Action, or any scheduler. ## How it works **1. Fix the entity list and the query template** Separate the two. One template — "education reform outcomes in " — crossed with an entity list is far easier to maintain and audit than hundreds of hand-written queries. **2. Window every run** Set `year_min` and `month_min` to the last successful run rather than filtering client-side. You retrieve less, pay for less, and get a cleaner set. **3. Paginate to exhaustion per entity** Walk `page` from `0` while `is_end` is false. A single query returns at most 1,000 results in total, so if an entity reaches that window, split it into narrower queries rather than paging further. **4. Diff on DOI against stored state** Persist every DOI you have reported. The run's output is the set difference, not the search results. **5. Log the empty runs too** A run that finds nothing is a record you need — it distinguishes "no new evidence" from "the job silently broke three weeks ago." ## The pipeline ```python theme={null} import os, time, requests API = "https://api.consensus.app/v1/search" HEADERS = {"x-api-key": os.environ["CONSENSUS_API_KEY"]} STRATEGY = dict( study_types=["rct", "meta-analysis", "systematic review"], exclude_preprints=False, # early signals often appear as preprints first page_size=200, ) def sweep(query, since): """Every record for one query since the last run.""" page, out = 0, [] while True: resp = requests.get(API, headers=HEADERS, params={ "query": query, "year_min": since.year, "month_min": since.month, "page": page, **STRATEGY, }) if resp.status_code == 429: # rate limited — back off and retry time.sleep(int(resp.headers.get("retry-after", 30))) continue resp.raise_for_status() body = resp.json() out.extend(body["results"]) if body["is_end"] or body.get("next_page") is None: break page = body["next_page"] if (page + 1) * body["page_size"] > 1000: # a query returns at most 1,000 results break return out def run(entities, template, since, seen_dois): new, counts = {}, {} for entity in entities: results = sweep(template.format(entity=entity), since) counts[entity] = len(results) for paper in results: doi = paper.get("doi") if doi and doi not in seen_dois: new[doi] = {**paper, "entity": entity} return new, counts # persist counts even when new is empty ``` > [!NOTE] > > Pagination past the first page requires a paid plan. `page_size` defaults to 20 and caps at 20 on Free, 300 on Pro and Teams, 750 on Deep, and 1,000 on Enterprise; a request over your ceiling is clamped rather than rejected, so read the `page_size` echoed back in the response. On every plan a single query returns at most 1,000 results in total, because `(page + 1) × page_size` can't exceed 1,000 — reach wider coverage with more queries, not more pages. ## Rate limits and backoff Build backoff in before you widen the entity list, not after. * **Pace to one search per second.** That is the limit on every plan except Enterprise, so build the pacing into the pipeline rather than firing entities in parallel. Honour `retry-after` and add jitter between entities. * A **429 mentioning your monthly search allowance** is a usage limit, not a rate limit — retrying will not help. Turn on additional usage in your account settings to keep searching past the included allowance. * Scheduled runs consume searches like any other request. A 121-entity sweep every six hours is roughly 484 entity-sweeps a day before pagination; size your plan for it. ## Reporting the delta Rank the new papers by whether they change a decision, not by citation count — a paper published this week has none yet. A useful report is five items with one line each and a link, plus an explicit "nothing notable this run" state that actually gets used. ## What to check before you trust it * **Alert on the job, not just the findings.** A silent pipeline looks identical to a quiet literature. Persist per-run counts and alert when a run returns zero results across *all* entities. * **Version the query template.** When you change it, the delta becomes incomparable to previous runs. Record the template version alongside each run. * **Watch for entity-name collisions.** Short or ambiguous entity names pull in unrelated work. Spot-check the highest-volume entities. * **Cache if you re-query.** Heavy users hit this fast; cache on the full parameter set with a short TTL. ## Related > [card] [Monitor new research about a drug, device, or brand](/use-cases/brand-and-device-monitoring) > > The same machinery, pointed at a commercial watchlist. > [card] [Build a reusable literature-review workflow](/use-cases/literature-review-skill) > > The interactive version of the same method. > [card] [Best practices](/use-cases/best-practices) > > Exhaustive sweep and diff-against-last-run in detail. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Write the search-strategy section of your methods Source: https://docs.consensus.app/use-cases/search-strategy-methods Turn a logged search into the reproducible methods paragraph reviewers ask for — accurate because it was recorded as it ran. ## What you get The paragraph that says exactly what you searched, with which filters, on what date, returning how many records — written from a log rather than reconstructed from memory weeks later. ## Who it's for Anyone publishing a review, meta-analysis, or scoping study. Also anyone who has been asked by a reviewer to "provide the full search strategy" and discovered they cannot. ## Why this is worth automating Search strategies are almost always written last, from half-remembered queries, after the interesting work is done. The result is a paragraph that does not reproduce the corpus it claims to describe — which reviewers increasingly check. If the agent logs each call as it runs, this section costs nothing and is actually true. ## What to log, as you go | Field | Why a reviewer wants it | | - | - | | Query text, verbatim | The only way anyone reproduces your set | | Every filter, exactly as sent | `study_types`, `human`, `year_min`, `sjr_max` all change the result set | | Date the search ran | Literature moves; a search has an as-of date | | Records returned per query | Pre-deduplication totals for the flow diagram | | Records after deduplication | Reported separately from the above | | Pages walked | Shows whether you exhausted the result set or took page one | | Any failed or retried call | A gap in coverage you must disclose | ## The prompt ```text theme={null} You are writing the search-strategy subsection of a methods section from the log below. Write prose, not a table — journals want a reproducible narrative. Cover, in this order: 1. The source searched, named precisely, with the date the search was run. 2. The query set. Give the queries verbatim; do not paraphrase them. 3. Every filter applied, with the value used and one clause on why. A filter with no stated rationale reads as arbitrary. 4. How records were deduplicated and on what key. 5. The counts: records identified, duplicates removed, records screened. 6. Any limitation: failed searches, truncated result sets, plan-imposed caps on returned results, or coverage restricted to a subset of the literature. Rules: - Use only the log. If a field is missing, write [NOT RECORDED] rather than reconstructing it — an invented search strategy is worse than an incomplete one. - Do not describe the search as comprehensive or exhaustive unless the log shows every query was paginated to completion. - State plainly that this was one search source and name any others used separately. Log: [PASTE THE AUDIT LOG] ``` ## Getting the filter rationale right Reviewers reject filters that look like convenience. Each one needs a defensible reason: * `study_types` — the designs that can answer the question, not the ones that returned pleasing results * `human=true` — scope of the review, stated in the question * `year_min` — tie it to an event, a guideline change, or a method's introduction, not to a round number * `sjr_max` — quality thresholds are contestable; say why the cut is where it is * `exclude_preprints` — a real methodological choice with a cost either way; justify the direction ## Audit requirements This use case *is* the audit log, rendered as prose. Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability), and keep the machine-readable table as a supplement — many journals now accept it as an appendix, which is stronger than the paragraph alone. ## What to check before you trust it * **Reproduce your own corpus from the paragraph.** If you cannot, a reviewer cannot either, and that is the entire test. * **Check for `[NOT RECORDED]`.** Each one is a real gap. Fill it by re-running, not by remembering. * **Do not claim comprehensiveness you did not achieve.** A truncated sweep described as exhaustive is a correctable error before submission and a serious one after. ## Related > [card] [Build a PRISMA-ready search corpus](/use-cases/prisma-corpus) > > The sweep that generates this log. > [card] [Screen abstracts against inclusion criteria](/use-cases/abstract-screening) > > Where the exclusion reasons come from. > [card] [Best practices](/use-cases/best-practices) > > Grounding rules and the audit-log format. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Check a supplier's performance claim against independent research Source: https://docs.consensus.app/use-cases/supplier-claim-check Test what a vendor says their material, tool, or process achieves against what the published literature independently reports. ## What you get A per-claim verdict on a supplier's technical assertions: independently supported, supported only under narrower conditions, contradicted, or unstudied — with citations and the specific conditions each finding applies to. ## Who it's for Technical evaluation teams, process and materials engineers assessing a new supplier, and anyone whose qualification programme is about to be sized around a vendor's datasheet. ## How it works **1. Split the datasheet into testable claims** "Higher throughput and better uniformity" is two claims with different evidence. Extract each as a proposition naming the material or system, the condition, the metric, and the comparator. **2. Pin down the comparator** "30% better" is meaningless without what it is better than. If the vendor has not said, that is itself a finding — record it and search against the conventional alternative. **3. Search the mechanism, not the brand** Trade names rarely appear in the literature. Search the underlying material, chemistry, or technique. This is the step that determines whether you find anything at all. **4. Run the negative pass** Search explicitly for null findings and failures with this approach. Independent groups publish disappointments that vendors do not mention. **5. Check the conditions match yours** A claim independently confirmed under conditions unlike your process is not confirmed for you. State the delta. ## The API calls ```bash theme={null} # independent evidence for the mechanism curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=graphene-enhanced thermal interface material conductivity measured performance" \ --data-urlencode "domain=mat,eng,phys" \ --data-urlencode "page_size=50" # negative pass curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=graphene thermal interface material no improvement limitations contact resistance" \ --data-urlencode "domain=mat,eng,phys" \ --data-urlencode "page_size=50" ``` ## Grading each claim | Verdict | Meaning | | - | - | | **Independently supported** | Multiple independent groups report comparable results under comparable conditions | | **Supported, narrower** | Confirmed, but only under conditions narrower than claimed. State which | | **Vendor-only** | The only supporting work is authored or funded by the supplier | | **Contradicted** | Independent work reports the opposite or finds no effect | | **Unstudied** | No independent literature. Not the same as false — say what you searched | The **vendor-only** row does most of the work here. A claim supported exclusively by the supplier's own publications is not independently verified, and that distinction rarely survives a datasheet review otherwise. ## What to check before you trust it * **Check author affiliations on supporting papers.** Supplier-authored or supplier-funded work is legitimate evidence but is not independent, and the difference matters for a qualification decision. * **Unstudied is not disproven.** A genuinely novel material will have thin independent literature. Treat it as a reason to test rather than a reason to reject. * **Coverage is patchy in this literature.** Much applied engineering work sits in closed proceedings; a thin result is weak evidence of absence. * **This informs qualification, it does not replace it.** The literature tells you what to test and what to watch for, not whether the part works in your process. ## Related > [card] [Compare candidate materials against published test data](/use-cases/materials-comparison) > > Evaluate the alternatives on the same footing. > [card] [Substantiate a product claim with published evidence](/use-cases/claim-substantiation) > > The same discipline applied to your own claims. > [card] [Best practices](/use-cases/best-practices) > > Filter ladder and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Swap a hand-maintained paper corpus for full-text search Source: https://docs.consensus.app/use-cases/swap-a-paper-corpus Replace a curated PubMed subset you maintain yourself with a weekly-refreshed full-text index, without rebuilding the infrastructure. ## What you get Your agents stop reasoning over a stale slice of abstracts and start reasoning over full-text findings, with quality filters and a link back to every source. And nobody on your team maintains a corpus pipeline any more. ## Who it's for Engineers who own "the literature service" at an AI-for-science company — the one who inherited a scraped or licensed PubMed subset and now owns its refresh, its gaps, and its quality problems. ## What it replaced at Owkin [Owkin](https://consensus.app/home/community-voices/owkin/) builds AI tools for biopharma drug discovery. Its flagship product, K Pro, is an Agentic AI Scientist orchestrating specialized agents across clinical trial design, patient and population decisions, and early portfolio choices. Early versions of K Pro ran on a curated collection of PubMed abstracts. Serviceable, but it left agents reasoning from titles and summaries rather than full-text findings, with limited filtering for study quality and no built-in defense against hallucination. > "We went from maintaining a collection of PubMed abstracts to 200M+ full-text articles updated weekly, with hallucination safeguards built in. It's exactly the kind of best-in-class building block we want to compose with, rather than rebuild." > > — **Lucas Brechot**, Senior Product Manager, AI Products, Owkin | | Result | | - | - | | Paper coverage | **25x** increase | | Index refresh | **Weekly** | | Time to ship | **Days**, against weeks or months of custom building | ## Why abstracts are the wrong foundation **1. The findings are not in the abstract** Effect sizes, conditions, populations, and limitations live in the body of a paper. An agent restricted to abstracts produces answers that are confident and shallow — the worst combination in a high-stakes domain. **2. A self-maintained corpus decays** Someone has to keep it fresh, handle schema drift, and explain the gaps. That work is invisible until it breaks, and it is never the thing your product is differentiated on. **3. Quality filtering has to be programmatic** Human researchers instinctively weigh recency, citation count, and journal reputation. Agents need those as parameters — `year_min`, `citation_min`, `sjr_max` — or they treat a 1998 case report and a 2025 meta-analysis as equivalent. **4. Grounding needs a link, not a claim** Every result carries a `doi` and a `url` back to the paper. That is what turns "the model says" into "this paper says", which is the difference that matters to your users. ## Making the swap Point your retrieval layer at the [MCP server](/consensus-mcp) or [`GET /v1/search`](/api-get-started), and map your existing corpus filters onto the parameters that replace them: | What you maintained | What replaces it | | - | - | | A recency cutoff on your ingest | `year_min`, `month_min` | | A journal allowlist | `sjr_max`, `journal_name`, `publisher_name` | | A study-design tag you derived yourself | `study_types`, `human`, `controlled` | | A minimum-quality heuristic | `citation_min`, `sample_size_min` | | Your own full-text extraction | `full_text_chunks` | | A domain restriction | `domain`, `medical_mode` | > [!NOTE] > > Full-text excerpts (`full_text_chunks`) come back on paid plans and cover open-access papers only. Pagination also needs a paid plan. See [rate limits](/use-cases/best-practices#rate-limits-and-result-caps). ## What to check before you trust it * **Run your old corpus and the new one side by side** on the same queries before cutting over. You are looking for regressions on your niche, not average quality. * **Check the filters actually narrow the way your old pipeline did.** A corpus restriction you enforced at ingest becomes a per-request parameter, and it is easy to forget one. * **Confirm coverage on your specific subfield.** Full-text excerpts currently cover open-access papers, so a closed-access niche will look thinner than it is. * **Keep the DOI.** If your product cites papers, the `doi` and `url` are what make the citation checkable downstream. ## Related > [card] [Add a citable evidence layer to an agent you already ship](/use-cases/add-an-evidence-layer) > > The same integration, framed as an addition rather than a replacement. > [card] [Benchmark literature providers against your own ground truth](/use-cases/benchmark-retrieval) > > How to decide whether the swap is actually an upgrade. > [card] [Best practices](/use-cases/best-practices) > > Rate limits, grounding rules, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Turn a syllabus into a recommended reading list Source: https://docs.consensus.app/use-cases/syllabus-reading-list Extract course topics and learning outcomes from a syllabus, then produce a curated reading list of recent research with summaries and discussion questions. ## What you get A ready-to-distribute reading list: 15–25 recent peer-reviewed papers grouped under your course's own topic sections, each with a clickable link, a plain-language summary written for students, and a discussion question tied to a stated learning outcome. Plus an audit log showing exactly which searches produced it. Consensus ships this as a packaged skill. The workflow below is what it does, so you can run it as-is, adapt it, or build it into your own tooling. ## Who it's for Faculty refreshing a course, curriculum designers, and corporate training teams who need current research rather than last year's list. ## Before anything else: check the tool is attached The workflow is gated on Consensus actually being callable — not installed, not authorized elsewhere, but present in this conversation. Verify before parsing the syllabus, and if it is missing, stop and ask the user to attach it with the **+** button or by tagging `@Consensus`. Do not fall back to web search or model knowledge for paper discovery; the whole point is verifiable citations. ## Phase 1 — Parse the syllabus Extract two things from the uploaded file (PDF, DOCX, image, or pasted text): 1. **Course topics** — the ordered list of units. These become the reading list's sections. 2. **Learning outcomes** — the course-level goals. These drive the discussion questions. If the syllabus states no explicit outcomes, infer 3–5 from the course description and topic list. Group closely related topics — "Protein Structure" and "Protein Function" become one section. Aim for **6–12 sections**: enough granularity to be useful, not so many that the document fragments. Present the grouping and confirm it before searching. ## Phase 2 — Search per section Write the full query plan first and record its total count. The phase is complete only when every planned query has produced an actual response and appears in the audit log. A zero-result response counts as complete; an intended call does not. **Query design is where this workflow succeeds or fails:** * Include the core topic **plus an applied angle**. For a biochemistry course with a nutrition focus, don't search `enzyme kinetics` — search `enzyme kinetics food processing applications`. * If the course has an obvious applied domain, weave it into *every* query. That is what surfaces papers bridging theory and application, which make far better supplementary reading than narrow primary research. * Keep queries to **4–8 words**. Longer queries dilute relevance. | Section | Query | | - | - | | Nucleic Acids | `nucleic acid structure function food safety` | | Amino Acids | `amino acid nutrition metabolism health` | | Enzyme Function & Kinetics | `enzyme kinetics food processing applications` | | Lipids & Membranes | `lipids biological membranes lipoproteins nutrition` | | Energy Metabolism in Disease | `energy metabolism diabetes insulin resistance` | Set `year_min` to the current year minus one by default. Run searches **sequentially** — send, wait, confirm, then send the next. See [rate limits](/use-cases/best-practices#rate-limits-and-result-caps) for the per-tier ceiling. **Selection criteria, in priority order:** 1. Relevance to the course topic 2. Reviews and meta-analyses over narrow primary research — they give students a broader entry point 3. Higher citation counts 4. Clear connection to the course's applied domain Select **1–3 papers per section**, targeting 15–25 total. ## Phase 3 — Write summaries and questions Base every summary only on what the search returned — title, abstract, metadata. If a result lacks enough context for a meaningful summary, say so rather than inventing detail. **One-sentence summary**, written for undergraduates. Define technical terms in parentheses; the summary should make a student want to read the paper. > **Good:** "This review maps how different diets — Mediterranean, Nordic, and vegetarian — reshape the types of fat molecules circulating in your blood, with implications for heart disease risk." > > **Bad:** "This paper reviews lipidomic profiles across dietary interventions and their cardiometabolic implications." — too jargon-heavy for the audience. **Discussion question** tying the paper to a learning outcome, pushing past recall into apply, analyze, or evaluate. > **Good:** "If dietary fat quality can reshape your lipoprotein lipidome, what does this suggest about the biochemical basis for dietary guidelines recommending unsaturated over saturated fats?" > > **Bad:** "What did the authors find?" — pure recall. ## Phase 4 — Build the document An editable Word document, not a PDF, so the instructor can cut sections into their own materials: * Title block: course title, generation date, publication-year range * A short introduction stating the readings came from Consensus results in this session * A **Course Learning Outcomes** section * Numbered papers under each confirmed topic heading * Per paper: clickable title linked to its full Consensus URL, authors, journal, year, plain-language summary, discussion question * A **Search Audit Log** Render the finished document to page images and inspect every page for overflow, clipping, broken links, and orphaned headings before delivering. ## Audit requirements Follow the [grounding rules](/use-cases/best-practices#grounding-and-auditability). For this workflow specifically, the audit log carries one row per section: section, query, filters as sent, papers found, papers returned, papers selected, status. Then the totals — queries sent, papers returned, papers cited — plus any detected plan cap and any failure. Report the same numbers in your closing message, not just in the document: > **Search summary:** Ran 12 queries across 10 sections. Consensus returned 34 papers. Selected 18 for the list. No failures. > > **Sections with limited results:** "Nitrogen Metabolism" returned only 1 relevant paper — worth supplementing manually. ## What to check before you trust it * **Check for padding.** A section with three recommendations where the field published little is the failure mode. Sparse sections should be flagged, not filled. * **Read the summaries as a student would.** If they read like abstracts, the audience instruction was ignored. * **Verify the level.** Recent frontier papers are often unsuitable for an introductory course regardless of relevance. * **Confirm access** before assigning anything students must read. Add `open_access=true` when students lack broad institutional access. * **Non-English courses:** search in English since Consensus indexes English-language papers, and note the course language in the document. ## Related > [card] [Build a reusable literature-review workflow](/use-cases/literature-review-skill) > > The deeper research version, with framework selection and search budgets. > [card] [Find NIH grants for a research idea](/use-cases/grant-finder) > > Another packaged skill built on the same search discipline. > [card] [Best practices](/use-cases/best-practices) > > Rate limits, grounding rules, and the five primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Build a drug target validation dossier Source: https://docs.consensus.app/use-cases/target-evidence-tiers Human genetics, in vitro, animal, and clinical evidence for one target in a single pass — including what contradicts it. ## What you get A target dossier organized by evidence tier, with a strength grade per tier, an explicit disconfirming-evidence section, and a named next experiment. The ratio between preclinical and clinical volume is itself part of the finding. ## Who it's for Discovery and translational teams deciding whether a target justifies a program, and reviewers pressure-testing that decision. ## The prompt ```text theme={null} You are a translational biology researcher building a target validation dossier. Target: [TARGET] Indication: [INDICATION] Search each evidence tier separately using Consensus. Report them separately — do not merge tiers into a single narrative. Tier 1 — Human genetics Loss-of-function and gain-of-function variants, GWAS associations, Mendelian randomization. Does human genetic variation in this target move the phenotype, and in which direction? Tier 2 — In vitro and mechanistic Cell models, biochemical assays, pathway position. What does modulating the target actually do, and in what system? Tier 3 — Animal models Knockout and knockdown phenotypes, pharmacological modulation, which species and which model. Note where the model is a poor proxy for the human disease. Tier 4 — Human observational Expression, biomarker association, natural-history cohorts. Tier 5 — Human interventional Any trial of an agent hitting this target or its pathway, in this or any indication. Include failures and terminations. For every tier, report: the number of papers found, what they show, the consistency of the direction of effect, and a strength grade of strong, moderate, weak, or absent. Then, as separate sections: Disconfirming evidence — search explicitly for contradictory findings, failed trials, non-replications, and negative results on this target. A dossier with no disconfirming section has not been done. If you genuinely find none, say what you searched. Safety and tractability — knockout phenotype severity, expression in tissues where modulation would be harmful, known adverse effects of agents hitting this pathway. Translational gap — state plainly whether human evidence exists or whether this is preclinical only. Report the ratio of preclinical to clinical papers. Finish with: the single experiment or dataset that would most reduce uncertainty about this target. Cite every claim with journal, year, and DOI. ``` ## How it works **1. Study-type filters are the tier boundary** `study_types=animal` and `non-rct in vitro` isolate preclinical work; `human=true` with `rct` and `cohort study` isolates the clinical tier. Running both and comparing volumes gives you the translational gap for free. **2. Disconfirming evidence is searched for, not waited for** Querying the target's name and indication returns confirmation. Finding the failures requires separate queries phrased toward failure — termination, non-replication, negative result. **3. Tiers stay separate through to the output** Merging tiers lets a large volume of in vitro work stand in for absent human evidence. Grading each tier independently prevents that. ## The API equivalent ```bash theme={null} # Tier 1 — human genetics curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=GPR75 obesity body mass index human genetic association loss of function" \ --data-urlencode "human=true" \ --data-urlencode "domain=med,bio" \ --data-urlencode "page_size=50" # Tier 3 — animal models curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=GPR75 knockout mouse adiposity phenotype" \ --data-urlencode "study_types=animal" \ --data-urlencode "page_size=50" # Disconfirming pass curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=GPR75 negative result failed replication no effect adiposity" \ --data-urlencode "page_size=50" ``` ## What to check before you trust it > [!WARNING] > > A literature dossier is an input to a target decision, not the decision. Keep the citation trail so a reviewer can check each tier's grade against its sources. * **Check the disconfirming section is real.** If it is empty or generic, the negative queries were not run. That section failing is the difference between a dossier and a pitch. * **Preclinical volume is not evidence strength.** Fifty in vitro papers and no human data is a weak target with a large literature. The ratio should be stated plainly. * **Watch for pathway substitution.** Evidence for a pathway is often presented as evidence for a specific target. Check that Tier 1 and Tier 5 are about the target itself. * **Gene-name synonyms matter.** Older aliases and gene-family names return different literature. Confirm the search covered them. ## Related > [card] [Create a biomarker evidence report](/use-cases/biomarker-report) > > The same tiered discipline, applied to a biomarker range. > [card] [Monitor new research about a drug, device, or brand](/use-cases/brand-and-device-monitoring) > > Keep watching the target once the program starts. > [card] [Best practices](/use-cases/best-practices) > > Fan-out, the filter ladder, and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Track how a field's terminology changed over time Source: https://docs.consensus.app/use-cases/terminology-shift Compare old and recent literature on the same concept so you search the terms the field actually used, not just the ones it uses now. ## What you get A mapping of old vocabulary to current vocabulary for your topic, a note on when the shift happened, and a search-term list covering both — so a review does not silently miss its own foundational literature. ## Who it's for Anyone entering an unfamiliar field, and every systematic reviewer, where missing older work is a methodological failure rather than an inconvenience. ## Why this matters more than it sounds "Gut flora" became "gut microbiome". "Manic depression" became "bipolar disorder". "Mental retardation" became "intellectual disability". Search only the current term and the foundational decade of a literature is invisible to you — and nothing in your results indicates it is missing. This is one of the few review failures that produces no error signal at all. ## How it works **1. Run the same concept in two eras** One search with `year_max` set before the suspected shift, one with `year_min` set after. Same concept, deliberately different windows. **2. Extract the vocabulary from each window** Pull the recurring noun phrases from titles and abstracts in each set. You are comparing how the field named the thing, not what it concluded. **3. Find the crossover** Narrow the window year by year to locate when the new term overtook the old. That date is worth reporting — it tells later readers where your search boundary sits. **4. Check whether the concepts are actually identical** Terminology changes sometimes accompany a definitional change. "Gut flora" and "gut microbiome" are close to interchangeable; other renamings narrowed or widened what counts. Say which case applies. **5. Build the combined term list** Produce a single list with both vocabularies and an era note on each, ready to paste into any search. ## The API calls ```bash theme={null} # older window curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=intestinal bacteria host metabolism" \ --data-urlencode "year_max=2010" \ --data-urlencode "page_size=100" # current window curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=gut microbiome host metabolism" \ --data-urlencode "year_min=2021" \ --data-urlencode "page_size=100" ``` Beyond vocabulary, the two windows also reveal **conclusion shifts** — where older papers reached different findings, signalling a paradigm change or accumulating evidence — and **methodological evolution**, where observational work gave way to trials and then to meta-analyses. Both tell you how mature the evidence base is. ## What to check before you trust it * **Confirm the old term actually returns results.** A term that returns nothing may be wrong rather than obsolete; check it against a paper you know exists. * **Watch for terms that were never synonyms.** Adjacent-but-different concepts appearing in the old window will pull irrelevant work into your review if you treat them as equivalent. * **Note regional and disciplinary variation.** Some vocabularies split rather than shift, and both remain current in different communities. * **Record the term list in your methods.** It is part of the search strategy, and a reviewer who spots a missing synonym will ask. ## Related > [card] [Build a reusable literature-review workflow](/use-cases/literature-review-skill) > > Era-gated searches as one step in a full review. > [card] [Build a PRISMA-ready search corpus](/use-cases/prisma-corpus) > > Where a missed synonym does the most damage. > [card] [Best practices](/use-cases/best-practices) > > Filter ladder and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Check whether a finding has human evidence yet Source: https://docs.consensus.app/use-cases/translational-readiness Separate cell and animal work from human studies for a claim, and report the ratio — the number that tells you how translated a finding really is. ## What you get A clear split of the evidence into preclinical and human tiers, with counts for each, the strongest study in each, and an explicit verdict on whether the claim has ever been tested in people. ## Who it's for Discovery and translational teams deciding whether a finding is ready to build on, investors doing scientific diligence, and anyone evaluating a claim that sounds better-established than it is. ## Why the ratio is the finding A large preclinical literature reads as strong evidence. Forty mouse studies and zero human trials is a very different situation from four of each — but a naive literature summary presents both as "well studied". Making the tier split explicit is the entire point of this workflow. ## How it works **1. Run the same question at two tiers** One search restricted to preclinical designs, one restricted to human studies. Identical concept, deliberately different populations. **2. Use study types as the tier boundary** `study_types=animal` and `non-rct in vitro` isolate preclinical work. `human=true` with `rct`, `cohort study`, and `non-rct observational study` isolates the clinical tier. **3. Report both counts, always** Even when one is zero. Especially when one is zero. **4. Check whether the human work tested the same thing** Human studies often measure a surrogate rather than the outcome the preclinical work implied. A human trial of a biomarker is not a human trial of the disease outcome — say which you found. **5. Look for failed translation explicitly** Search for trials that ran and did not replicate the preclinical finding. This is the most informative and most frequently skipped search. ## The API calls ```bash theme={null} # preclinical tier curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=senolytic clearance of senescent cells lifespan extension" \ --data-urlencode "study_types=animal,non-rct in vitro" \ --data-urlencode "page_size=100" # human tier curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=senolytic therapy clinical trial outcomes older adults" \ --data-urlencode "human=true" \ --data-urlencode "study_types=rct,cohort study" \ --data-urlencode "page_size=100" # failed translation curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=senolytic trial did not replicate preclinical finding null result" \ --data-urlencode "human=true" \ --data-urlencode "page_size=50" ``` ## Reporting it | Tier | Papers | Strongest study | Verdict | | - | - | - | - | | In vitro | 34 | … | Consistent effect | | Animal | 41 | … | Consistent across two species | | Human observational | 3 | … | Association only | | Human interventional | 0 | — | **Never tested in people** | The bottom row is the output. Everything above it is context. ## What to check before you trust it * **Zero human studies may mean zero, or may mean your query missed them.** Confirm with a different phrasing before reporting an absence — human trials often use a compound name where preclinical work uses a mechanism. * **Watch the surrogate outcome.** "Tested in humans" for a biomarker change is not "tested in humans" for the clinical outcome. * **Species matters.** Effects in short-lived model organisms translate especially poorly for anything involving lifespan or ageing. * **Absence of failed trials is not evidence of success.** Negative results are under-published, so a clean record may reflect publication bias. ## Related > [card] [Build a drug target validation dossier](/use-cases/target-evidence-tiers) > > The full tiered dossier this check is one section of. > [card] [Grade a set of papers with your own evidence rubric](/use-cases/evidence-grading-rubric) > > Score what you find on a consistent scale. > [card] [Best practices](/use-cases/best-practices) > > Filter ladder and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona. # Find which journals to submit to Source: https://docs.consensus.app/use-cases/where-to-submit Identify the journals actually publishing work like yours, with their quartile, so you target a realistic venue first. ## What you get A shortlist of journals that publish your kind of work, each with how many closely-related papers they carry, their SCImago quartile, and representative papers you can read to judge fit. ## Who it's for Researchers choosing a submission target — especially early-career authors, and anyone publishing into a field they are new to. A mis-targeted submission costs a months-long rejection cycle before you learn anything. ## How it works **1. Search as if you were looking for your own paper** Query the specific contribution, not the broad topic. "Machine learning in healthcare" returns the wrong venues; "transformer models for ICU mortality prediction" returns the ones that would review you. **2. Sweep deep, not wide** Paginate. The venue distribution in the top 20 results is noisier than across 200, and specialist journals surface further down. **3. Tally journals with quartiles** Count `journal_name` across the deduplicated set, carrying `sjr_best_quartile` and `publisher_name`. Frequency tells you where this conversation happens; quartile tells you how hard the door is. **4. Split by recency** Rank once over five years and once over the last eighteen months. A journal that published heavily in 2019 and nothing since has moved on from the topic. **5. Read three papers per candidate** Fit is about framing and method, not keywords. The representative papers are the point of the exercise. ## The API call ```bash theme={null} curl -G "https://api.consensus.app/v1/search" \ -H "x-api-key: $CONSENSUS_API_KEY" \ --data-urlencode "query=transformer models ICU mortality prediction electronic health records" \ --data-urlencode "year_min=2022" \ --data-urlencode "page_size=200" ``` Then aggregate client-side: ```python theme={null} from collections import Counter venues = Counter() quartile = {} for paper in results: name = paper.get("journal_name") if not name: continue venues[name] += 1 quartile[name] = paper.get("sjr_best_quartile") for name, n in venues.most_common(15): print(f"{n:>3} Q{quartile.get(name, '?')} {name}") ``` Add `open_access=true` if your funder mandates OA, and `sjr_max=1` to see only the top-quartile subset — but look at the unfiltered list first, or you will miss the specialist venue that is the actual best fit. ## What to check before you trust it * **Frequency is not acceptance probability.** The journal publishing the most papers in your area may also reject the most. Use the list to build a realistic ladder — a reach, two solid targets, a fallback. * **Check scope pages by hand.** Journals shift scope, launch sections, and close them. The literature lags those changes by a year or more. * **Watch for a topic sitting mostly in conferences.** In fast-moving computational fields the venue that matters may not be a journal at all, and journal-only tallies will mislead you. * **Quartile is a blunt instrument.** SCImago quartiles vary by subject category; a Q2 in a narrow specialist category can be more selective than a Q1 in a broad one. ## Related > [card] [Find a real citation for every claim in a draft](/use-cases/ground-a-manuscript) > > Get the manuscript itself defensible before you submit it. > [card] [Track how a field's terminology changed over time](/use-cases/terminology-shift) > > Make sure you are searching the terms this field actually uses. > [card] [Best practices](/use-cases/best-practices) > > Exhaustive sweep and the rest of the primitives. > [card] [All use cases](/use-cases/gallery) > > Browse the gallery by persona.