Skip to content

Getting Started

Get AutoDocs MCP Server running with your AI assistant in under 5 minutes!

🚀 Quick Setup

Step 1: Install AutoDocs

uv tool install autodoc-mcp

pip install autodoc-mcp

git clone https://github.com/bradleyfay/autodoc-mcp.git
cd autodoc-mcp
uv sync --all-extras

Step 2: Test the Installation

# Verify installation
autodoc-mcp --version

# Test the server (should show FastMCP startup screen)
autodoc-mcp

You should see output like: 

🚀 FastMCP Server Starting...
📊 Available MCP Tools: 8
✅ Server ready on stdio transport

Press Ctrl+C to stop the test server.

Step 3: Configure Your MCP Client

Choose your AI assistant and follow the configuration:

# AutoDocs is already globally available after installation
# Start it in your session:
autodoc-mcp

Add to Cursor settings (Cmd+, → Extensions → Rules for AI → MCP Servers): 

{
  "mcpServers": {
    "autodoc-mcp": {
      "command": "autodoc-mcp",
      "env": {
        "CACHE_DIR": "~/.cache/autodoc-mcp"
      }
    }
  }
}

Add to claude_desktop_config.json: 

{
  "mcpServers": {
    "autodoc-mcp": {
      "command": "autodoc-mcp",
      "env": {
        "CACHE_DIR": "~/.cache/autodoc-mcp"
      }
    }
  }
}

Step 4: Test with Your AI Assistant

Start a conversation with your AI assistant and try these test prompts:

Test Commands

  1. "What packages are available in this project?" (Uses scan_dependencies tool)

  2. "Tell me about the FastAPI package with its dependencies" (Uses get_package_docs_with_context tool)

  3. "What's in the AutoDocs cache?" (Uses get_cache_stats tool)

🎯 What to Expect

First Request (Cache Miss)

  • Response Time: 3-5 seconds (fetching from PyPI)
  • Context: Primary package + 3-8 relevant dependencies
  • Caching: All responses cached for future use

Subsequent Requests (Cache Hit)

  • Response Time: < 100ms (instant from cache)
  • Context: Same comprehensive context as before
  • Performance: Lightning fast with preserved quality

📊 Success Indicators

✅ Working Correctly

  • AI assistant acknowledges AutoDocs tools are available
  • Package queries return structured documentation
  • Dependencies are automatically included in context
  • Cache builds up over time (check with "cache stats" query)

❌ Common Issues

  • "Tool not found" errors: MCP configuration problem
  • Network timeouts: Check PyPI connectivity
  • Empty responses: Verify package names are correct

See Troubleshooting for detailed solutions.

🧠 Understanding the Context System

AutoDocs provides three levels of context:

Level 1: Primary Package

{
  "name": "fastapi",
  "version": "0.104.1",
  "summary": "FastAPI framework, high performance...",
  "key_features": ["Automatic API docs", "Type hints", "Async support"],
  "main_classes": ["FastAPI", "APIRouter", "Depends"]
}

Level 2: Runtime Dependencies

{
  "runtime_dependencies": [
    {
      "name": "pydantic",
      "why_included": "Required by fastapi",
      "summary": "Data validation using Python type hints"
    }
  ]
}

Level 3: Context Metadata

{
  "context_scope": "smart (3 deps)",
  "total_packages": 4,
  "token_estimate": 15420,
  "performance": {
    "total_time": 0.89,
    "cache_hits": 1,
    "cache_misses": 3
  }
}

🎨 Usage Patterns

For Development Work

Ask your AI assistant: - "How do I create a FastAPI app with database models?" - "What are the key differences between FastAPI and Django?" - "Show me how to set up authentication with these packages"

For Learning & Exploration

Ask your AI assistant: - "What does this project depend on and why?" - "Explain the relationship between these frameworks" - "What are the main features of package X?"

For Debugging & Troubleshooting

Ask your AI assistant: - "This error mentions Pydantic - what might be wrong?" - "How do I fix version conflicts with these dependencies?" - "What's the proper way to configure package X?"

⚡ Performance Tips

Maximize Cache Efficiency

  • Ask about common packages first (FastAPI, Django, requests)
  • Use consistent package names (avoid typos that miss cache)
  • Let AutoDocs choose dependencies (smart scoping is optimized)

Optimize for Your Workflow

  • Set environment variables for your preferred settings
  • Use project-specific cache directories for isolation
  • Monitor cache stats to understand performance patterns

📈 What's Next?

Now that AutoDocs is working:

  1. Explore MCP Tools: Learn about all 8 available tools
  2. Customize Configuration: Tune performance with configuration options
  3. Advanced Usage: Discover power-user features in the API Reference

🆘 Need Help?