MCP tools

KAOS is MCP-native: every tool is typed, annotated, and exposable over the Model Context Protocol. Each package ships a kaos-*-serve entry point that serves its tools to any MCP client (Claude Code, Codex, …); kaos-mcp bridges any runtime to MCP over stdio or streamable HTTP.

Tools per package

Approximate counts (the authoritative list is generated per package — see below):

Package	MCP tools	Serve command
kaos-core	~10	— (use kaos-mcp)
kaos-content	~17	—
kaos-llm-client	~7	kaos-llm-serve
kaos-llm-core	~32	kaos-llm-core-serve
kaos-agents	~14 (+7 retrieval)	kaos-agents-serve
kaos-pdf	~8	kaos-pdf-serve
kaos-office	~18	kaos-office-serve
kaos-tabular	~17	kaos-tabular-serve
kaos-source	~30	kaos-source-serve
kaos-web	~45	kaos-web-serve
kaos-nlp-core	~17	kaos-nlp-serve
kaos-nlp-transformers	~7	kaos-nlp-transformers-serve
kaos-ml-core	~11	kaos-ml-serve
kaos-graph	~17	kaos-graph-serve
kaos-citations	3	kaos-citations-serve
kaos-ui	4 (read-only)	—
kaos-mcp	0 (pure bridge)	kaos-mcp serve
kaos-names	0	—

Note

Counts are intentionally shown as approximate here — the single source of truth is each package’s tool registry (the monorepo’s scripts/inventory.py enumerates every tool, resource, and schema). Prose on this site never hard-codes exact counts elsewhere, so the docs can’t drift from the packages.

How tools are exposed

• Every tool is a KaosTool with typed inputs (schema derived from type hints) and annotations (readOnlyHint, destructiveHint, …) that let an agent’s permission policy decide what to auto-allow.
• kaos-mcp mounts a whole runtime: kaos-mcp serve (stdio) or --http (streamable HTTP, which requires an explicit auth token).
• Connect a serve command to an MCP client to make KAOS tools available inside it.

Enumerate tools live

uv run --with kaos-pdf kaos-pdf-serve --help     # or list-tools via kaos-mcp