What it is

When an AI agent is asked to build something physical — a robot, a sensor rig, an embedded project — it eventually needs real parts: motors, drivers, sensors, fabrication services. Today that means the agent either hallucinates a plausible-sounding part, or a human has to go find one manually. Xibrary is the layer in between: a structured, queryable index the agent calls directly, the same way it might call any other tool.

It is a sourcing and discovery layer only. Xibrary does not design products, and it does not place orders on its own — it answers "where do I get this part, legally, at what price, how fast" with data an agent can act on without a human re-verifying it first.

Why every result carries a confidence tier

Every listing and every compliance record in Xibrary is tagged with an explicit confidence tier — VERIFIED, AUTO_CLASSIFIED, or UNKNOWN. An agent (or a person) should never treat UNKNOWN export-control or pricing data as authoritative. This is the thing that makes Xibrary's data safe to act on autonomously, instead of just another source that needs a human to double-check it.

For AI agents / developers

Connect via MCP (recommended) or plain REST:

MCP endpoint:  https://xibrary-api.fly.dev/mcp   (Streamable HTTP, stateless)
Tools:         search_parts, get_part_specs

REST:          GET https://xibrary-api.fly.dev/parts?category=electronics&subcategory=motors
               GET https://xibrary-api.fly.dev/parts/{id}

OpenAPI spec · llms.txt · llms.txt (mirror)

For manufacturers and suppliers

As AI agents increasingly do the choosing instead of a person browsing search results, being represented in the structured data those agents query is what determines whether they find you at all. Xibrary indexes public catalog data, structured and attributed back to you, at no cost during this early build phase. If you received an email about this and want to see what a listing would look like, or want to be included, reply to that email — we're doing this with suppliers' knowledge and input, not without it.

From the blog

We write about AI agents, the Model Context Protocol, agentic commerce, and how Haven Command builds — including what game design teaches us about designing for autonomous agents. Read the blog →