Elicit's Two-Front Expansion: Multiplayer Research Workspace + Evidence API/MCP Backend
Elicit — the academic evidence-synthesis platform — is now executing on both sides of the research-tool bifurcation simultaneously: it is building up into a collaborative destination workspace while also selling its evidence engine as a backend for other people's agents. This makes it the clearest live test of the dual-layer thesis in Destination vs. Backend: The Research-Tool Market Bifurcates — and Both Layers Converge on MCP.
Front 1 — the workspace/destination layer. The August 5, 2026 launch of the "Research Agent" repositioned Elicit from literature-review assistant to a decision environment: it "synthesizes evidence across not only scientific literature but a breadth of data sources such as trials, patents, regulatory databases, structured biology databases, the web, and your own uploaded data," and outputs "fully cited reports, tables, analyses, visualizations, slides, and documents," with a "transparent methods trail and sentence-level citations." Elicit claims its Smartest mode covers 76.7% of key decision considerations vs. 68.8% for Claude Opus 5 Max on its internal benchmark (Elicit changelog, Aug 5 2026). Then on September 1, 2026 it went multiplayer:
"AI has enabled a single researcher to do more work than ever before. Today, we're launching a set of features that make Elicit more collaborative and give teams a shared environment where researchers and AI agents can work together." — Elicit changelog, Sept 1 2026
The Sept 1 release ships Skills (teams "turn repeatable research processes into Skills that anyone in the organization can use"), Collaborative Sessions (teammates comment on artifacts, request edits, interact with the agent, version history), and Shared Projects (persistent home for sessions, files, sources, artifacts). Elicit's own framing: "These updates are important steps toward making Elicit a home for collaborative research, where the work done today becomes useful context for the work that comes next."
Front 2 — the backend/evidence layer. On July 15, 2026 Elicit publicly launched its API and MCP server, exposing three endpoints: a Search API (138M+ papers and ClinicalTrials.gov records, semantic or keyword, filters by year/journal quartile/study type, retracted papers excluded by default), a Reports API (question in, fully cited markdown/PDF/DOCX report out, parallelizable), and a Systematic Review API (full search→screening→extraction→report flow, "each decision recorded, so the review is auditable end to end"). All three are also served over MCP, so "Elicit can return cited answers directly in Claude, ChatGPT, Gemini, or any MCP-compatible client" (Elicit blog, Jul 15 2026).
The technical claim underneath: "Elicit was built for this problem before agents even existed. Since 2018 we've worked on an approach called factored cognition: instead of one opaque model doing all the work, evidence synthesis runs as discrete building blocks for search, screening, extraction, and reporting." And the retrieval benchmark claim: "Our API had the best search results (paper recall) when compared against four of the most popular search tools on BioASQ benchmark."
Customer pull for the backend is agent-native: a VC fund runs scheduled jobs and coding agents against the API to produce audio research briefings because, in the founder's words, much of today's AI-generated research is still "vibes on the internet"; a life-sciences investment firm cut ~20 hours of upfront research per asset; a healthcare agency replaced a raw PubMed integration with Elicit as the literature layer of an internal pharma-insights engine (~80% prep-time reduction expected).
Pricing (Elicit pricing page, fetched 2026-09-19): Basic free (unlimited search across 138M+ papers, limited agent usage); Pro $49/user/mo billed annually ($588/yr — systematic review screening of 5,000 papers, 20 extraction columns, API access); Scale $169/user/mo annually ($2,028/yr — 5x usage, figure extraction, real-time collaboration, admin panel); Enterprise custom (40,000-paper screening, PRISMA-grade accuracy, SSO/SAML, single-tenancy, no training on your data, unlimited API). Since June 30, 2026, limits moved "from workflow-based limits to a monthly usage pool" — the same usage-pool pricing shift seen across the market (changelog).
What this means for the Hey, Lefty map
- Elicit now occupies both layers at once: destination (multiplayer workspace, shared projects, skills) and backend (API/MCP evidence provider for other agents). It is the vertical (academic/pharma) benchmark for the strategy Consensus was already tracking in Market Map & Positioning: Where Does Hey, Lefty Fit?.
- For Hey, Lefty, Elicit is simultaneously a candidate MCP source (its Search/Reports/Systematic Review endpoints are exactly the kind of evidence backend an orchestrator would call) and a competitor for the workspace seat in science-adjacent research.
- The "Skills" construct — reusable, team-defined research procedures — is emerging as the market's packaging unit for expertise (Elicit's Skills, Moonshot's 9 financial Skills in Kimi for Financial Services: Moonshot Sells Vertical Research at $7–$104/Month, Wired via MCP to S&P Global and Wind). Whatever Hey, Lefty calls its orchestration recipes, the category is converging on this vocabulary.