Research Paper Search API - Academic Papers, Studies, Citations

Search academic research papers and scientific studies with study design filters, or ask a research question and get a cited, evidence based answer, through the Consensus API with your own API key, as one clean dataset row per paper or answer.

Consensus is an AI academic search engine over more than 400 million scholarly records, built on Semantic Scholar, OpenAlex, PubMed, arXiv and its own crawl, and enriched with study type, sample size, population, journal quartile and a one line takeaway per paper. Its REST API on api.consensus.app serves paper search and the Threads research agent. This actor wraps both: give questions or keyword queries and get deduplicated rows you can download as JSON, CSV or Excel, schedule, or feed into literature reviews, evidence tables, RAG pipelines, AI agents, Google Sheets and internal research tools. Bring your own API key.

BYOK, not live-tested: every path, parameter and enum comes from the Consensus OpenAPI document (docs.consensus.app/openapi.json), and the API was probed with an invalid key (each wrapped route answers 401, an invented route answers 404), but the actor has not been run against a live Consensus key.

What it does

• Paper search: papers for each research question or keyword query, ranked by relevance, with title, authors, year and publish date, journal, publisher, volume, pages, DOI, Consensus link, citation count, influential citation count, study type, sample size, number of studies in a meta-analysis, population (human or animal), countries of study, study duration, institutions and the abstract and takeaway.
• Study filters: year and month range, study types (RCT, meta-analysis, systematic review, cohort, case-control, cross-sectional and 17 more), human or controlled studies only, minimum sample size, minimum citations, study duration, SCImago journal quartile, peer reviewed only, open access only, clinical guidelines only, fields of study, countries of study, publishers and a preferred journal.
• Medical mode, relevance scores and full text excerpts: search only top medical journals and guidelines, add the semantic score, or add up to three query relevant full text passages per paper (paid Consensus plans).
• Research questions with cited answers: the Consensus research agent reads about 20 papers (Pro) or about 50 (Deep) and returns a markdown answer whose markers cite numbered papers and their verbatim quotes, plus counts of papers retrieved, screened and included. Several questions run side by side.
• Research skills: literature review, citation snowballing, find citations and verify citations workflows in Deep mode.
• Follow-ups and history: continue a thread with follow-up questions, read any earlier answer by thread ID, and list your threads with an answer preview.
• Pay only for results: refused, empty and unfinished lookups are recorded as a free row that says why, never charged.

How to get an API key

Sign in to Consensus, open the API and MCP dashboard and create a key. The free plan includes 30 calls a month; Pro and Teams include 500 and Deep 2,000, and paid plans can turn on metered overage at 0.05 USD per call. Paste the key into the API key field, or set the DATA_API_KEY secret. The key is sent only to api.consensus.app in the x-api-key header, never in the URL, and is removed from everything the actor logs.

Input

Field	Modes	Description
mode	all	search (default), research, followUp, getThread or listThreads
queries	search	Questions or keyword queries, one per line, up to 50
resultsPerQuery	search	Papers per query, 1 to 1,000 (default 20)
medicalMode, includeSemanticScore, includeFullTextChunks	search	Medical corpus, relevance score, full text excerpts
questions, researchBrief	research, followUp	Research questions (up to 20) and one long brief
agentMode, skill	research, followUp	pro or deep; a named skill (Deep only)
paperIds, collectionIds	research, followUp	Papers or Library collections to focus on
maxWaitSeconds, includeTrace	research, followUp	Wait per answer (default 600 s), agent trace
threadId	followUp	The thread to continue
threadIds, interactionId	getThread	Threads to read, optionally one earlier answer
titleQuery	listThreads	Title substring
yearMin, yearMax, monthMin, monthMax	search, research	Publication date range
studyTypes, fieldsOfStudy	search, research	Study designs and academic fields
humanOnly, controlledOnly, excludePreprints, openAccessOnly, clinicalGuidelinesOnly	search, research	Yes or no filters
sampleSizeMin, citationMin, durationMinDays, durationMaxDays	search, research	Numeric filters
journalQuartileBest, journalQuartileWorst	search, research	SCImago quartile range, 1 is top
countries, publisherNames, journalName	search, research	Country of study, publishers, preferred journal
maxResults	all	Row ceiling for the run (default 100)
requestsPerMinute	all	Pacing, default 50
apiKey	all	Your Consensus API key (or the DATA_API_KEY secret)

Example: randomized trials and meta-analyses on creatine and memory since 2015, top quartile journals only.

{
    "mode": "search",
    "queries": ["does creatine improve memory"],
    "studyTypes": ["rct", "meta-analysis"],
    "yearMin": 2015,
    "humanOnly": true,
    "journalQuartileWorst": "1",
    "resultsPerQuery": 50
}

Example: two research questions answered with citations.

{
    "mode": "research",
    "questions": [
        "What does the evidence say about vitamin D supplementation and depression in adults?",
        "Is intermittent fasting more effective than continuous calorie restriction for weight loss?"
    ],
    "agentMode": "pro",
    "excludePreprints": true
}

Example output

An illustrative paper row, shaped like the documented response (values are an example, not real data).

{
    "mode": "search",
    "modeLabel": "Paper search",
    "recordType": "paper",
    "query": "does creatine improve memory",
    "rank": 1,
    "id": "10.1000/example.2023.001",
    "title": "Effects of creatine supplementation on memory in healthy individuals: a systematic review and meta-analysis",
    "authors": ["A. Example", "B. Example"],
    "year": 2023,
    "publishDate": "2023-03-14",
    "journalName": "Example Nutrition Reviews",
    "publisherName": "Example Press",
    "doi": "10.1000/example.2023.001",
    "url": "https://consensus.app/papers/example",
    "citationCount": 120,
    "influentialCitationCount": 9,
    "studyType": "meta-analysis",
    "sampleSize": 225,
    "studyCount": 10,
    "populationType": "human",
    "journalQuartile": 1,
    "isPreprint": false,
    "takeaway": "Creatine supplementation may improve memory performance in healthy individuals, especially older adults.",
    "abstract": "...",
    "fullTextChunks": null,
    "found": true,
    "retrievedAt": "2026-10-22T12:00:00.000Z",
    "note": null
}

A research answer row carries threadId, interactionId, threadTitle, threadUrl, status, agentMode, skill, question, answer (markdown with markers such as [2.1] for quote 1 of paper 2), citations (each with n, paperId, title, authors, year, journal, doi, url and snippets of marker, quote and section), citedPapers, papersRetrieved, papersScreened, papersIncluded and, when asked for, agentTrace. A thread row carries threadId, threadTitle, status, preview, interactionCount and lastActivityAt.

FAQ

What is the Consensus API?

It is the REST API of Consensus, an AI search engine for peer reviewed research. GET /v1/search returns papers with study metadata and filters, and the Threads endpoints run the Consensus research agent, which writes cited answers. It authenticates with an x-api-key header and is documented at docs.consensus.app.

Do I need my own Consensus API key?

Yes. This is a bring-your-own-key actor: it calls Consensus with your key, under your plan and terms. A missing or rejected key ends the run cleanly with a status message that explains what to change, and nothing is charged.

How do I find randomized controlled trials or meta-analyses on a topic?

Choose Paper search, enter the topic as a question, and pick rct, meta-analysis or systematic review under Study types. Add Human studies only, a minimum sample size or a journal quartile to narrow it further.

How is a research question different from a paper search?

Paper search returns the ranked papers themselves. A research question hands the question to the Consensus research agent, which searches, screens and reads papers and returns a written answer where every claim cites a numbered paper and a verbatim quote. It takes from about one to ten minutes per question.

What if a research answer is not ready in time?

Each question waits up to the Wait for answer limit (600 seconds by default). A run still going is saved as a free row with its thread ID; run Get thread with that ID later to collect the answer. Keep the actor run timeout above the wait limit.

How many Consensus calls does a run use?

Search is billed by Consensus at one call per 100 papers, rounded up, so this actor asks for pages of up to 100. A research thread is billed by the work the agent does. Polling, Get thread and List threads are reads. The run log reports the exact request count.

What are the plan limits?

Paging past the first page of a search, full text excerpts, Deep mode and skills need a paid plan; the page size is capped at your plan's maximum, and no query returns more than 1,000 papers. The API allows about one request a second per plan.

What do I pay?

Pay per result: 0.005 USD per paper, 0.02 USD per finished research answer and 0.005 USD per listed thread. Refused, empty and unfinished lookups are never charged. Your Consensus plan is billed separately by Consensus, including any overage at 0.05 USD per call.

Pricing

Event	Price	Charged for
paper-result	0.005 USD	Each paper returned by a search
research-answer	0.02 USD	Each finished research answer with citations
thread-record	0.005 USD	Each earlier research thread listed

Keyword map

Research paper search API, academic search API, scientific papers API, scholarly literature search, peer reviewed studies API, systematic review search, meta-analysis finder, randomized controlled trial search, evidence based answers API, citation API, literature review automation, research agent API, RAG for research papers, Consensus API, Consensus app API, academic research data.

Terms and data use

Consensus describes its API as a way to bring Consensus results into your internal tools, workflows and agents, and its Terms of Service grant a personal, non transferable, non sublicensable licence. Treat the results as for your own internal research use: do not resell or republish them as a dataset without written permission from Consensus. The docs encourage caching on the full parameter set with a short refresh time so new literature still appears, and ask you to keep each excerpt with its paper's DOI and link. Paper abstracts and excerpts remain the copyright of their publishers. This actor is not affiliated with or endorsed by Consensus.