Bench — provenance-tracked AI co-pilot for clinical and biotech research
A desktop research co-pilot aimed at clinical and biotech PIs who use ChatGPT to draft grant text, summarize papers, and spot statistical problems in their data — but can't let an unsourced number leave their laptop. Every AI-generated sentence points back to the notebook, figure, or PDF it came from.
Principal investigators and senior scientists in biotech and translational medicine who write grants and manuscripts with AI
- Citation-traceable draft paragraphs pinned to local notebooks, figures, and PDFs so every claim is auditable
- Local-first, model-agnostic runtime that works with on-prem Llama, OpenAI, or Claude under one workflow
- Built-in review skills for stats-integrity, figure-or-code consistency, and unsourced-number detection
- Export to DOCX and LaTeX with provenance metadata embedded for journals and IRBs
Trending 489-star repository this week signals labs are actively rejecting chat-only AI tools and demanding local, auditable research workbenches with provenance baked in.
Strong signal: Springer study evaluates ChatGPT as autonomous grant-writing tool and Polish automation paper details full grant workflow with GPT; Elicit reports 2M+ researcher users across academia and industry.Assessing ChatGPT's potential as an autonomous grant writing tool ↗Pricing | Elicit: The AI Research Assistant ↗
Crowded: Elicit (2M users), Scite, Consensus, SciSpace, MACg, ResearchRabbit, plus Anthropic's Claude Science and the exact 'Open Science' open-source repo (★489 trending) already shipping local-first provenance workbench on Tauri/MCP.GitHub - ai4s-research/open-science ↗Open Science - Open-Source AI Workbench for Scientists ↗Scite: AI for Research ↗
Proven willingness to pay: Elicit publishes explicit tiered pricing, Scite sells enterprise plans, MACg targets life-sciences budgets; biotech PIs control grant overhead that routinely pays for premium SaaS ($20–50/seat/mo equivalent).Pricing | Elicit: The AI Research Assistant ↗Launch of MACg AI Medical Writing, Research, and Reference Management ↗
Provenance/auditability is a durable regulatory pull (FDA/EMA AI guidance, reproducibility crises, funder data-sharing mandates), and local-first aligns with growing data-sovereignty norms in clinical research.Scite: AI for Research ↗
Non-trivial: requires deep PDF/figure parsing, citation graph, local LLM orchestration, notebook integrations (Jupyter/R), and PDF provenance linking; the referenced Open Science repo is already building this stack on Tauri+MCP, raising the bar.GitHub - ai4s-research/open-science ↗