# KaaS

> Self-hosted, open-source knowledge-base compiler turning scattered notes into a queryable Markdown wiki via a 4-phase LLM pipeline. Exposes wiki via MCP for AI agents.

- **Category:** Knowledge Management
- **Licence:** MIT
- **Latest version:** v0.1.0-rc.2
- **Open source:** oss
- **Languages:** Typescript, Go, Python
- **Platforms:** Docker, Linux
- **Tags:** SQLite, Markdown, Mcp, Wikis, Ai Agents, Knowledge Base, Knowledge Management, LLM, Rag, Documentation, Typescript, React, Python, Golang, Self Hosted, Openai Api, Note Taking, Model Context Protocol, Personal Knowledge Management, Claude code, Second brain, Openai compatible

## Description

KaaS (Knowledge as a Service) compiles raw content — pasted text, uploaded files, or fetched URLs — through a four-phase LLM pipeline (Extract, Classify, Write, Index) into human-readable Markdown articles organized as a wiki. Unlike typical retrieval systems that break text into embedded fragments, KaaS produces structured, cited articles you can read, edit, and version with Git. The compiled wiki is queryable via a web UI with streaming cited answers, or through MCP (Model Context Protocol) for integration with coding agents like Claude Code and Codex. It runs as a single Docker image bundling a Go backend, React frontend, and Python AI daemon, using SQLite by default with optional MySQL. Configuration is via TOML file with environment variable overrides for secrets. The project is MIT-licensed and developed by Bybit Exchange.

## Links

- [Homepage](https://github.com/bybit-exchange/kaas)
- [Website](https://bybit-exchange.github.io/kaas-doc/)
- [Docker](ghcr.io/bybit-exchange/kaas)
- [GitHub Repository](https://github.com/bybit-exchange/kaas)
- [MCP Integration Guide](https://bybit-exchange.github.io/kaas-doc/guides/mcp-integration.html)
- [CLI Reference](https://bybit-exchange.github.io/kaas-doc/reference/cli.html)
- [Configuration Reference](https://bybit-exchange.github.io/kaas-doc/reference/configuration.html)

---
Source: https://selfhost.directory/project/kaas