AI and MCP inside ManuFind, explained
How ManuFind's AI actually works — and how you can plug your own assistant into it safely.
"AI" is on every product page these days, so it's worth being specific about what it means inside ManuFind. It isn't a chatbot bolted on the side — it's wired into your documents, and it's built to be both useful and safe.
Answers you can trust
When you ask a question, ManuFind first works out what you actually need — a search across your library, a lookup of structured data, or a drafted document — and routes it to the right tool. Every answer cites the documents it came from, so you can verify it instead of taking it on faith. A security layer screens each request for prompt injection and filters the output, and the AI only ever sees what the person asking is allowed to see.
Open by design: MCP
ManuFind speaks the Model Context Protocol (MCP) — an open standard for connecting AI assistants to real tools. In plain terms: you can point an assistant like Claude or Cursor (or your own agent) at ManuFind and let it actually do things — search documents, pull a BOM, draft a quote — through 73 governed tools. You're not locked into our chat box; bring the assistant your team already likes.
Built for builders too
Prefer to integrate directly? There's a REST API with 140+ endpoints, an OpenAPI spec and Postman collection, scoped mf_live_ keys, and webhooks — so ManuFind fits into whatever you already run.
A two-minute walkthrough
Here's what it looks like in practice. An engineer connects Claude to ManuFind with a scoped API key — one setting, no custom code. They ask: "What material did we spec for the Henderson job, and is there a newer revision?" Behind the scenes the assistant calls ManuFind's search and document tools, ManuFind checks what that engineer is allowed to see, and the answer comes back with the exact drawing and revision cited. One click opens the source document — no folder spelunking, no asking around, no guessing.
Safety isn't an afterthought
Every AI action, every agent, and every API key obeys the same role-based access control as your people. Data stays isolated per customer, AI runs on Amazon Bedrock inside our cloud, and your documents are never used to train public models. Powerful, but never at the expense of security. For the full technical picture, read our whitepaper: Secure AI by architecture.