MCP (Model Context Protocol) is an open standard for connecting AI assistants to tools and data, instead of every product inventing its own bespoke integration. An MCP server exposes a defined set of tools; any MCP-compatible client - Claude, your own agent, whatever you're running - can call them.
It's scoped deliberately narrow: read-only knowledge retrieval, never a control plane for managing your Yew instance itself. The tools that exist today:
Right now, the MCP server authenticates the same way the rest of Yew does - your logged-in session. That works well for connecting your own MCP-compatible client to your own instance. What it doesn't support yet is an external AI tool connecting on someone else's behalf without an interactive login - that needs a real API-key permission model, which is still being built. We'd rather tell you that plainly than let the word "MCP" imply more than what's actually running.
Once that lands, the plan is two consumers of the same server: Yew's own conversational assistant (also not built yet), and external AI tools you already use, plugged directly into your knowledge without needing Yew's own interface at all.