Undermind

Undermind is an AI co-researcher purpose-built for academic literature search and scientific discovery, founded by quantum physics PhDs from MIT. Rather than keyword search, Undermind deploys an autonomous agent that conducts successive, adaptive searches of scientific databases: following citation trails, adjusting strategy based on prior results, and reading hundreds of papers to build comprehensive coverage of a research area. The agent asks clarifying questions before beginning, then generates annotated reports with inline citations, thematic summaries, evidence hierarchies, contradictions, and gaps across the literature. Coverage draws from Semantic Scholar (234M+ papers across PubMed, arXiv, ACM, IEEE, bioRxiv, and institutional repositories). Free tier allows 5 searches per month with abstract-only analysis; Pro unlocks unlimited searches with full-text analysis. Best fit: academic researchers, pharmaceutical scientists, and R&D teams conducting deep literature reviews; graduate students building field command; and enterprise research teams integrating structured literature evidence into AI workflows. Undermind draws on the Semantic Scholar API (234M+ papers) as its foundational data layer, providing access to PubMed, arXiv, ACM, IEEE, and bioRxiv content through a single research interface. An enterprise API enables institutional integration into broader AI research stacks: GSK embedded Undermind as the research-evidence layer of its internal AI Scientist platform, a documented deployment demonstrating meaningful integration depth for pharmaceutical research workflows at scale. Research area monitoring sends alerts within the Undermind interface when new relevant papers are published. Named integration gaps for individual researchers: no Zotero or Mendeley native connector (manual reference copy required), no Overleaf or LaTeX integration for manuscript workflows, no official MCP server for agent-to-agent integration as of Q2 2026, and no Slack or Teams delivery for research alerts. Undermind follows a publicly listed freemium model. Free: $0/month, 5 searches per month with abstract-only analysis. Pro: $16/month billed annually (20% saving over monthly) or $20/month billed monthly, unlimited searches, full-text analysis where available, and unlimited report chats. Team: custom pricing for 5+ researchers with centralized billing, collaboration features, and priority support. Enterprise: custom pricing for institutions and pharmaceutical companies, including API access and dedicated support. The major confirmed enterprise deployment is GSK, where over 1,000 scientists use Undermind daily as the literature-evidence layer of its internal AI Scientist platform. Undermind is not suited for fast-answer or non-academic research needs. The depth-first citation-trail methodology is optimized for comprehensive literature command and takes materially longer per query than fast-answer tools. For AI-generated answers with citations without deep-search overhead, Consensus (from $9.99/month) provides quick claim-level answers across papers. For systematic review workflows requiring PRISMA-compliant screening and structured data extraction, Rayyan (free basic tier) or Elicit ($10/month) provide dedicated systematic review pipelines. For open-web research beyond peer-reviewed literature, Gemini Deep Research (5 reports/month free, full access via Google AI Pro at $19.99/month) or ChatGPT Deep Research ($20/month) cover broader source territory. For multilingual literature coverage, Undermind is primarily English-language; SciSpace (from $12/month) offers stronger multilingual capabilities for international research workflows. Undermind (YC S24) has built strong independent validation across academic channels: Tooliverse aggregates 180+ user reviews at a consensus score of 9.3/10; Comparateur-IA shows 4.7/5 from 68 user reviews; and a peer-reviewed product evaluation was published in a health sciences library journal and indexed on PMC/NCBI and ResearchGate. The GSK enterprise deployment adds an independently confirmed NPS of 63 (95th percentile for enterprise software) from a survey of 130+ scientists, reported independently by BankInfoSecurity, GovInfoSecurity, and CIO. No MCP server has been published as of Q2 2026. G2 shows 0 verified reviews, reflecting an academic and pharmaceutical adoption channel rather than commercial software procurement platforms. Annual billing at $16/month saves 20% over the $20/month monthly rate.