Academic research agent over a 138M-paper index that runs reports and systematic reviews, citing AI claims at sentence level.
Evidence teams running structured literature work rather than one-off questions: the site describes systematic review screening, extraction tables and Routines for scheduled agent runs, and states a single pass can find up to 1,000 relevant papers and analyse up to 20,000 data points. It fits reviewers who must defend each claim, because Elicit states every AI-generated claim carries sentence-level citations back to the source. The REST API and MCP server make it workable for groups wanting the same screening and extraction inside their own pipelines.
It is not a cheap or casual tool: the free Basic plan quantifies only search (138M+ papers) and leaves Research Agent and Report usage as 'limited', API access starts at Pro $49 per user per month, and results per API search request are capped at 300 on Pro, 500 on Scale and 10,000 on Enterprise. Screening is not a replacement for a second human reviewer on its own numbers: Elicit reports 93.6% recall but 62.8% specificity across 58 published systematic reviews, and the docs note 'fast' screening wrongly excludes more papers that met your criteria. The databases behind the 138M-paper index are never named on the site, so corpus coverage cannot be audited before you buy.
Pros
Cons
Its own index, which the docs describe as over 138 million academic papers; the pricing page says search across 138M+ papers is unlimited even on the free plan. The underlying databases are not named. Paper search can be restricted to a PubMed corpus, trial search returns ClinicalTrials.gov records, and journal quartile data comes from SCImago Journal Rank.
Yes, more than most vendors. It reports 94-99% extraction accuracy against 128 vetted gold-standard answers, 94% on an internal LLM-judged extraction evaluation, 96% on Methods, Participants and Interventions, and 99.4% in a VDI/VDE customer check of 1,511 data points. It also cites a randomised noninferiority preprint reporting a 1.0% hallucination rate across 5,100 data points. The figures come from different test sets.
Its published numbers argue against that. Elicit's screening evaluation across 58 published systematic reviews reports 93.6% recall and 62.8% specificity, meaning many irrelevant papers are screened in and roughly one relevant paper in fifteen is missed, with half the misses traced to badly written screening criteria. The docs also warn that fast screening wrongly excludes more papers that met your criteria.
Yes. REST API v2 covers paper and trial search, reports, systematic reviews, agent sessions, routines, library imports and usage, and there is an MCP server using OAuth 2.0 rather than an API key. Access requires Pro ($49/user/month) or above. Results per search request are capped at 300 on Pro, 500 on Scale and 10,000 on Enterprise, with a shared 100 requests/minute per-IP burst limit.
Facts sourced from the vendor’s own site, last verified 2026-10-10.
The Weekly Source
One email a week. New listings, repriced tools, and the ones we would skip - including this one if its pricing changes.
