Koog Documentation MCP Server
A Model Context Protocol (MCP) server that provides vector-based RAG search capabilities over Koog documentation using ChromaDB.
Setup
- Create and activate virtual environment:
python -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate- Install dependencies:
pip install -r requirements.txt- Clone the koog git repository and remember the path of your working copy.
- Add to your Q CLI MCP configuration:
q config mcp add koog-docs python3 koog-mcp-server.py [koog-repo-path]If you omit koog-repo-path, it will default to '../koog'
Or manually add to your MCP config file: ``json { "mcpServers": { "koog-docs": { "command": "python3", "args": ["koog-mcp-server.py"], "cwd": "/path/to/this/directory" } } } ``
Database Initialization
The vector database will be automatically created and populated on first use:
- ChromaDB creates a persistent database in
./chroma_db/ - All markdown files from
../koog/docs/docsare processed and indexed - Documents are chunked for optimal retrieval
- Vector embeddings are generated using ChromaDB's default model
Note: First startup may take a few moments while the documentation is indexed.
Usage
Once configured, you can ask questions about Koog in Q CLI:
- "How do I create a basic agent in Koog?"
- "What are the different types of tools available?"
- "How does memory work in Koog agents?"
- "Show me examples of structured output"
The server will search through all Koog documentation files and return relevant sections with context.
Features
- Vector-based semantic search using ChromaDB
- Automatic document chunking for better retrieval
- Persistent vector database with embeddings
- Searches all markdown files in the Koog docs directory
- Returns relevant sections with similarity scores
- Extracts document titles and file paths for reference
