cloudparts

Elicit

An AI research assistant over 138 million academic papers and 545,000 clinical trials that searches, screens and extracts data from studies and writes reports with sentence-level citations, built for literature and systematic reviews.

What it is
Retrieval-augmented generation (RAG)
Type
Commercial product
Made by
Elicit Research, PBC
License
Proprietary
Facts checked
13 October 2026
Website

What it is

Elicit retrieves from scientific literature rather than the open web. Its search runs semantic or keyword queries over about 138 million papers and ClinicalTrials.gov records, filters by year, journal quartile and study type, and leaves out retracted papers by default. On top of search it splits evidence synthesis into separate steps, an approach its makers call factored cognition: screen each paper against criteria, extract data into table columns (including from tables and figures), then write a report in which each claim cites the sentence it came from. Each step is evaluated on its own, and every screening decision keeps its reasons and supporting quotes so a systematic review can be audited in PRISMA style.

The free Basic plan includes unlimited search, summaries and chat with papers. Plus ($11 a month, billed yearly) adds clinical-trial search and exports; Pro ($39 a month, billed yearly) adds a systematic-review workflow that screens up to 5,000 papers, Routines and API access; Enterprise screens up to 40,000 papers and extracts 40 columns. Industry pricing is higher, for example $49 a month for Pro. Since July 2026 the Research Agent, Reports and systematic reviews draw on one monthly usage pool that grows with the plan.

The 2026 Research Agent widens the sources to patents, regulatory and biology databases, the web and uploaded files such as spreadsheets or omics data, and produces cited reports, tables, analyses and slides. The same capabilities are available through a REST API at https://elicit.com/api/v2 and an MCP server at https://elicit.com/api/mcp for Claude, ChatGPT and other agents.

Why teams pick it

  • Retrieval over scholarly sources, not web pages: 138 million papers plus clinical trials, with retracted papers excluded by default
  • Auditable steps: screening reasons, per-criterion scores and supporting quotes for every decision, and sentence-level citations in reports
  • Data extraction into tables, including from figures and tables in papers, for reviews of hundreds or thousands of studies
  • Usable from other agents through an API and an MCP server, with OAuth for MCP clients

Watch out for

  • API and MCP access need the Pro plan or above; search requests are capped per plan and all endpoints share 100 requests a minute per IP, with a 5-minute block if exceeded
  • Agentic work draws on a monthly usage pool, so heavy Research Agent or review use means a higher tier or paid extra usage
  • A hosted, closed service: “no training on your data by default” is listed as an Enterprise feature
  • Answers are only as complete as Elicit’s corpus and the full text it can reach; papers outside it have to be imported, for example from Zotero or as PDFs

Alternatives

What all retrieval-augmented generation (rag) have in common →Compare side by side →
KapaCommercialA managed retrieval-augmented generation service for technical companies: it keeps an index of your docs, tickets, forums and code in sync and answers questions from it with citations, as an “Ask AI” widget, Slack or Discord bot, support-form deflector, API or MCP server.

How it fits with other technologies

Search papers through the Elicit API (Pro plan or above) · bash

curl --fail-with-body https://elicit.com/api/v2/search/papers \
  -H "Authorization: Bearer $ELICIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query":"effects of sleep deprivation on cognitive performance","maxResults":10}'

# Or give Claude Code the MCP server (sign in with OAuth via /mcp)
claude mcp add --transport http elicit https://elicit.com/api/mcp

See it in a real stack

No architecture uses Elicit yet. Drop it into the playground and wire it up yourself.Open playground →
Where it came from2018–2026

Elicit grew out of Ought, a non-profit research lab whose mission was making sure progress in machine learning helps people reach good answers to hard questions. From 2018 the team worked on factored cognition: breaking reasoning into small, checkable steps instead of trusting one opaque model. Elicit applied that to research, starting by automating evidence synthesis: finding, screening and extracting data from academic papers. Andreas Stuhlmüller and Jungwon Byun lead it as CEO and COO.

To let Elicit raise money and scale up, in September 2023 Ought’s independent board sold its intellectual property to a new public benefit corporation, Elicit, which had raised $9 million. Most of Ought’s staff moved with it, Ought kept an equity stake, and the new company took on Ought’s mission.

In 2026 Elicit moved from fixed workflows toward agents. In July it opened its search, reports and systematic-review building blocks to other agents through an API and MCP server, and moved plans to a shared monthly usage pool; in August it launched the Research Agent for high-stakes decisions in science, with a benchmark of drug-development decisions; and in September it added Routines that keep a piece of research up to date.

Why it’s built this way

Separate, validated steps instead of one model answer
Under factored cognition, search, screening, extraction and reporting each run as a block that can be evaluated on its own, so the blocks can be recombined for a task without losing rigor.
Cite the exact sentence
Researchers need to check claims against the source, so reports and extractions carry sentence-level quotes and every screening decision keeps its reasons.
A public benefit corporation instead of a non-profit
Ought’s board judged that a company could raise the funding to scale Elicit while sharing Ought’s mission, and that the equity stake was also a prudent deal for the non-profit.

Timeline

  1. 2018

    Work on factored cognition begins at Ought

  2. 2023

    Ought spins off Elicit as a public benefit corporation with $9 million raised (September)

  3. 2026

    Elicit API and MCP server launched (July)

  4. 2026

    Research Agent and BioDecisionBench launched (August)

  5. 2026

    Routines keep research current (September)

Sources7 · facts checked 13 October 2026

Improve this page on GitHub