Elicit

A purpose-built AI research assistant spun out of Ought, a nonprofit ML research lab, and incorporated as a public benefit corporation in 2023. Elicit automates the most labor-intensive phases of scientific literature review — search, screening, data extraction, and synthesis — with a hard constraint on accuracy: every AI-generated claim is anchored to a sentence-level citation in the source paper. Used by over 2 million researchers across academia, pharma, and policy. For a firm, the primary use case is technology diligence: rapidly mapping the scientific literature behind a target company's core claims (quantum error correction, novel semiconductor materials, directed-energy systems, etc.) without commissioning a full expert review.

Semantic search across 138M+ academic papers (Semantic Scholar, OpenAlex, PubMed, ACS) and 545,000 clinical trials from ClinicalTrials.gov — no exact keyword match required. Automated Research Reports synthesizing up to 80 papers per brief, customizable by paper selection and extraction columns. Full Systematic Review workflow (PRISMA 2020-compliant): search/upload, abstract screening, full-text screening, structured data extraction across up to 5,000 papers, and final report. Custom extraction columns let users pull any variable (e.g., sample size, methodology, performance metric) into a structured table. Chat with Papers for interactive Q&A on uploaded PDFs. Research Alerts for monitoring new publications on a saved query. Research Agents (Pro+) for broader landscape mapping including clinical trial data, regulatory documents, and press releases beyond academic literature.

AI core uses a compositional multi-model approach (including Claude integration) rather than a single end-to-end black-box model; multiple models cross-check answers and confidence is only surfaced when models agree, materially reducing hallucination. Sentence-level citations on every claim. Validated at 95% search recall, 97% abstract screening accuracy, 99% full-text screening, and 96% extraction across 994 Cochrane reviews. API: available — search papers and generate Reports via REST API (see docs.elicit.com). MCP server: confirmed and officially supported. All API functionality is available via an MCP server at https://elicit.com/api/mcp; authentication via OAuth 2.0; compatible with Claude Desktop, Claude Code, and any MCP-compatible client. This is a strong MCP implementation — full API parity, OAuth-secured, with setup guides and examples published at github.com/elicit/api-examples.

Freemium with four tiers (as of 2025–2026, per Elicit support docs and third-party reviews). Basic: free — unlimited paper search and summaries, 2 automated reports/month. Plus: ~$12/month (or ~$84/year) — 4 reports/month, export (CSV/BIB/RIS), deeper extraction. Pro: $49/month or $499/year — 12 reports or systematic reviews/month (144/year), unlimited high-accuracy extraction columns, Research Agents, full systematic review workflow, 10 concurrent Alerts. Team: $79/seat/month or $780/seat/year (2-seat minimum) — pooled report quota (20/month per user), admin panel, unified billing, priority support. Enterprise/Institutional: custom pricing — contact sales@elicit.com; SOC 2 report available under NDA; supports institutional SSO and custom workflows.

Data sources: Semantic Scholar, OpenAlex, PubMed, ACS, ClinicalTrials.gov. Reference management: Zotero integration for exporting citations. Export formats: CSV, BIB, RIS on paid plans. API + MCP server enable embedding Elicit search and report generation into external workflows and AI agents. No native CRM, Slack, or document-management integrations publicly documented. Integration breadth is narrow outside research tooling — designed as a standalone research layer, not a hub.

SOC 2 Type II certified (achieved as of late 2024, confirmed via official Elicit blog post). SOC 2 report available to enterprise customers under NDA. User-uploaded content and search data are not used to train third-party public models. Multi-model compositional architecture filters problematic content including retracted papers. Trust Center publicly available at elicit.com. GDPR compliance not explicitly confirmed in public documentation — not publicly disclosed. HIPAA status not publicly disclosed. HQ is US-based (Oakland/San Francisco, CA).

Founded inside Ought (a nonprofit ML research lab established in 2017 by Andreas Stuhlmüller); product development began ~2021; spun out and incorporated as an independent public benefit corporation in September 2023. HQ: Oakland/San Francisco, CA. Co-founders: Andreas Stuhlmüller (CEO) — PhD in brain and cognitive sciences from MIT, postdoc in computation and cognition at Stanford, focused on automating reasoning; and Jungwon Byun (COO) — BA Economics from Yale, former Head of Growth at Upstart (scaled revenue 5x), former management consultant at Oliver Wyman. Funding: $9M seed in September 2023 led by Fifty Years, with Basis Set Ventures and angel investors including Jeff Dean (Google Chief Scientist) and Thomas Ebeling (former Novartis CEO); $22M Series A in February 2025 co-led by Spark Capital and Footwork; total raised ~$31–33M per PitchBook. Headcount: ~44–50 employees. Notable customers: Formation Bio (pharma), Oregon State University (institutional license), and researchers across pharmaceutical R&D, government policy, and academic institutions.

Moderate. Elicit is a strong diligence acceleration tool for a deep-tech growth equity firm like a firm — it can rapidly surface the scientific literature underpinning a target's technology claims (quantum computing architectures, novel semiconductor processes, directed-energy systems, space propulsion, etc.) in hours rather than days, with citations that can be handed to a technical advisor for spot-checking. The systematic review workflow and custom extraction columns are directly applicable to structured competitive landscape mapping across a sector. The confirmed MCP server with full API parity is a significant plus for an AI-native firm building automated diligence workflows in Claude or similar environments. Fit is Moderate rather than Strong because Elicit operates exclusively on published academic and clinical literature — it cannot search patents, SEC filings, private company databases, news, or internal documents, and it does not touch deal flow, CRM, LP data, or any firm-confidential information. It is a research layer, not a firm intelligence platform. Value is highest during initial technology assessment and market mapping; it does not replace expert technical diligence or commercial data providers.

Covers only published academic papers and clinical trials — no patents, regulatory filings (FDA, FCC, ITAR), SEC documents, news, or proprietary data sources. Search sensitivity can miss relevant recent papers, particularly if recently published or behind paywalls (cannot bypass paywalls for full-text access). Primarily indexes Semantic Scholar, creating database coverage gaps vs. multi-source tools. Research Agents on Pro+ expand scope to clinical trials and press releases but do not reach full web coverage. Not suited for non-empirical research domains. Steeper learning curve than general-purpose chat tools for structured extraction workflows. Customer support is async-only on sub-Enterprise tiers. Enterprise SSO and custom workflow features require direct sales engagement — no self-serve enterprise.