CLI Commands

KaaS provides two command-line programs: kaas (Go backend) and kb-ai (Python AI engine).

kaas

kaas serve

Start the KaaS backend server. The server exposes the REST API, Web UI, worker pool, and (optionally) the MCP endpoint.

kaas serve [-f <config-path>]

Flag	Description	Default
-f	Path to the TOML config file	etc/kaas.toml

The server listens on the address specified in the config ([server] addr), defaulting to 0.0.0.0:8080.

Environment variable overrides — see Environment Variables for the full list (LLM_API_KEY, LLM_BASE_URL, LLM_MODEL, etc.).

---

kb-ai

The Python AI engine. Typically managed by the Go backend as a daemon, but can also be used standalone.

kb-ai compile

Run the 4-phase compilation pipeline (Extract → Classify → Write → Index) on queued raw content. This is an internal command invoked by the Go worker; it reads a JSON request from stdin and writes a JSON response to stdout.

kb-ai compile

No flags — input is passed via the daemon bridge protocol (stdin JSON).

---

kb-ai fetch-url

Fetch the content of a URL, extract readable text, and return it as structured data. Used by the backend when a user submits a URL for compilation.

kb-ai fetch-url

No flags — input is passed via the daemon bridge protocol (stdin JSON).

---

kb-ai chat

Handle a chat request against the compiled wiki. Performs LLM-iterative retrieval over the knowledge base and streams a cited answer. Internal command invoked by the Go backend's chat endpoint.

kb-ai chat

No flags — input is passed via the daemon bridge protocol (stdin JSON).

---

kb-ai distill

Run the full distillation pipeline on local files or directories: ingest text content into the knowledge base's raw/ folder, then compile it into wiki articles.

kb-ai distill <paths...> [--kb <kb-dir>]

Argument / Flag	Description	Default
paths	One or more files or directories to distill	(required)
--kb	Knowledge-base directory	./.kaas

Example:

kb-ai distill ./notes ./meeting-transcripts --kb ./data

---

kb-ai mcp

Start the MCP (Model Context Protocol) server, exposing the ask tool for querying the wiki.

kb-ai mcp [--stdio | --http] [--host <host>] [--port <port>] [--kb-dir <path>]

Flag	Description	Default
--stdio	Use stdio transport (agent spawns the process)	(default)
--http	Use streamable-http transport	—
--host	Bind host (HTTP mode only)	127.0.0.1
--port	Bind port (HTTP mode only)	8082
--kb-dir	Override the knowledge-base root directory	KAAS_KB_DIR env or ./data

Environment variables:

Variable	Purpose
KAAS_KB_DIR	Knowledge-base root (fallback when --kb-dir not given)
KAAS_MCP_TOKEN	Bearer token for HTTP transport authentication
LLM_API_KEY	LLM provider API key
LLM_BASE_URL	LLM provider base URL
LLM_MODEL	Model to use for chat

Example — Claude Code integration (stdio):

claude mcp add kaas -- kb-ai mcp --kb-dir /path/to/data

---

kb-ai daemon

Start the AI engine as a long-running daemon process. This is an internal command — the Go backend spawns it automatically and communicates over a multiplexed stdin/stdout protocol.

kb-ai daemon

Not intended for direct use. The daemon listens for JSON commands on stdin and responds on stdout.