MCP
MCP reference
Kai runs a local MCP server over stdio, so connected coding tools can read your project map — privately, on your machine.
Connect an agent
Auto-wire Kai MCP into an agent’s configuration. Your repository code is never modified.
kai mcp claude # Wire Claude Code (~/.claude.json)
kai mcp cursor # Wire Cursor (.cursor/mcp.json)
kai mcp all # Wire every detected agentCheck status
Show the local MCP server status, available tools, and per-agent connection state.
kai mcp list # Or: kai list mcp, kai list -m, kai list --mcpDisconnect an agent
Remove Kai MCP from an agent. Repository code, caches, and project maps are preserved.
kai remove -m claude # Or: kai remove --mcp claude, kai mcp remove claudeOptions:
--print previews the snippet, --dry-run simulates without writing, --force overwrites an existing entry, --project writes project-scoped config. Disconnects accept --dry-run to preview first.Tools
9 local tools your agent can call. Everything runs on-device.
| Tool | Does |
|---|---|
kai_project_info | Project name, root, and map state |
kai_search_code | Search files, symbols, and snippets (query, up to 20 files) |
kai_find_symbol | Find a symbol by name |
kai_get_dependencies | Imports of a file |
kai_get_context | Context package for a task (token budget, default 4000) |
kai_refresh_index | Rebuild the map now (incremental unless forced) |
kai_get_lsp_status | Language server state |
kai_plan_task | Task plan with file priorities |
kai_sync_graph | Sync the map after writes |