📝 TL;DR

Scispot is the governed operating system for sample-centric labs. We just shipped an MCP server with 27 tools so Claude, Cursor, Claude Code, and ChatGPT can act directly on your live lab data with the same role permissions and audit trail as a human in the UI.

Hey everyone 👋

I'm Satya, co-founder of Scispot. Today we're launching the Scispot MCP Server, the first MCP server built for the governed lab operating system that scaling biotech, diagnostics, and regulated CROs run on.

🎯 The problem

Labs that are growing past 20–500 people don't have an "ELN or LIMS problem"; they have a fragmentation problem. Samples live in one system, results in another, instruments in a third, methods in Google Docs, and the answer to "which lot did this come from" lives in a person's head.

That fragmentation has three consequences AI makes worse, not better:

• Shadow workflows. Scientists copy sample data into Claude, ask it to find anomalies, paste the answer back, and (hopefully) update inventory. QA can't audit it.
• Governance leaks. A general-purpose agent has no idea a CRO coordinator shouldn't see another sponsor's samples, or that a clinical reviewer's role limits what they can edit.
• AI initiatives stall. Leadership wants AI outcomes. The data foundation isn't ready. CAP/CLIA and GxP teams say no until it is.

📈 Scispot's solution

We built Scispot as the governed operating system for sample-centric labs: sample-first data model, RBAC, audit, approvals, and a single API surface across every workflow a lab actually runs. The MCP server makes that operating system addressable by any agent:

• 27 tools spanning sample search and lineage, structured tables, methods and protocols, freezer and inventory state, lab flows, and image retrieval — all consistent under one auth and audit model.
• RBAC at the tool layer. Every call resolves the calling user's role. A CRO coordinator can't see another sponsor's samples through the agent. This is what unblocks regulated buyers.
• Audit + approvals inherited. Every MCP action lands in the same activity stream as a UI action, with actor, timestamp, and diff. Sensitive writes route through the same approval steps. The agent isn't a back door for CAP/CLIA or GxP.
• OAuth 2.0 + PKCE on top of a Scispot. Compatible with Claude Desktop, Claude Code, Claude Web, Perplexity, Cursor, and ChatGPT.
• One system, not five. Most customers retire 3–5 disconnected tools (LIMS, ELN, freezer, inventory, spreadsheets) on the way in. MCP gets one auth model, not five integrations.

🚀 Traction

• Live in production with 100+ labs
• 27 production tools shipped, with usage metering

🙋 What customers are doing with it

• pharma R&D: retiring a GraphPad + Claude + manual reporting stack. Their go-to prompt: "summarize all experiments linked to samples where Reagent Lot is XX-007." Answered against governed data, not a stale CSV.
• process development labs (e.g. fermentation) : asking Claude to walk sample lineage and storage state: "which box, which rack, which plate, which lineage" without learning the Scispot menu hierarchy.
• translational teams: wiring LaTeX-based methods writing to live experiments, so the methods section assembles itself from the actual run record. The next reader is a sponsor or regulator.
• AI/ML & Informatics teams: treating MCP as the programmatic spine of an ML pipeline instead of maintaining yet another export script. Sample-first data model, no separate warehouse to keep in sync.

📚 Backstory

We started Scispot because we watched scaling biotechs, molecular diagnostics labs, and regulated CROs lose weeks to fragmentation; every minute saved on the bench got eaten back at the spreadsheet. When MCP shipped, the standard finally let us put AI inside the governed system instead of beside it. We shipped MCP to the first cohort in March 2026.

🚀 Our ask

1. If you run a sample-centric lab — scaling biotech, molecular diagnostics, or regulated CRO — try the MCP server. Beta opt-in: survey.scispot.io/mcp

2. Intros welcome to Heads of Lab Ops, Informatics, or QA at labs feeling IND / CAP-CLIA / sponsor pressure. Email me directly: satya@scispot.io.

   See it in context: top 5 use cases · press release.