MCP

The MCP server exposes the knowledge graph to AI agents via the Model Context Protocol. It is a wire surface: it constructs an actor from each connection, calls the Access Policy to resolve read level or check write authority, and serializes shaped responses for agent consumption.

Each MCP tool corresponds to an abstract operation on the graph — read a node, traverse references, search, write a node, and so on. The tool surface is designed for agent ergonomics: structured output, progressive disclosure, traversal-friendly responses where each result carries the next address.

MCP does not query the storage layer directly, and it does not enforce access on its own. The Access Policy is the enforcement point; MCP is a consumer of that policy, the same as REST or any other wire surface.

See src/server/design.md for the reference implementation: the specific tool list, tool schemas, and how the server wires actor construction.