Dedalus MCP Documentation Server

An MCP server for serving and querying documentation with AI capabilities. Built for the YC Agents Hackathon.

Quick Start (Local Development)

# Install uv package manager (same as Dedalus uses)
brew install uv  # or pip install uv

# Install dependencies
uv sync --no-dev

# Configure API keys for AI features
cp config/.env.example .env.local
# Edit .env.local and add your OpenAI API key

# Test
uv run python tests/test_server.py

# Run
uv run main

Related MCP server: Whoosh RAG MCP

Deploy to Dedalus

What Dedalus Needs

• pyproject.toml - Package configuration with dependencies

• main.py (root) - Entry point that Dedalus expects

• src/main.py - The actual MCP server code

• docs/ - Your documentation files

Deployment Steps

1. Set Environment Variables in Dedalus UI:

   • OPENAI_API_KEY - Your OpenAI API key (required for AI features)

2. Deploy:

dedalus deploy . --name "your-docs-server"

How Dedalus Runs Your Server

1. Installs dependencies using uv sync from pyproject.toml

2. Runs uv run main to start the server

3. Server runs in /app directory in container

4. Docs are served from /app/docs

Features

• Serve markdown documentation

• Search across docs

• AI-powered Q&A (with OpenAI)

• Rate limiting (10 requests/minute) to protect API keys

• Ready for agent handoffs

Tools Available

• list_docs() - List documentation files

• search_docs() - Search with keywords

• ask_docs() - AI answers from docs

• index_docs() - Index documents

• analyze_docs() - Analyze for tasks

Documentation

See docs/ directory for:

• Getting Started Guide

• Hackathon Information

• Deployment Guide

• Examples

License

MIT