mcp -docy

Docy is an MCP server that provides instant, self-hosted access to technical documentation. It lets LLMs query and retrieve content from documentation sources by crawling allowed sites with crawl4ai, then serves that content through the MCP interface for accurate, up-to-date guidance. Docy exposes tools to list available documentation sources, fetch a specific documentation page as markdown, and discover links from a documentation page, enabling you to build rich, context-aware documentation queries into your AI workflows. When used with Claude Code or MCP-enabled tools like VS Code, Docy helps your AI cite and reference documentation effectively, reducing reliance on stale or external fetches. To ensure Claude Code prioritizes Docy, add the recommended guidelines to your CLAUDE.md file and use the provided tool prompts to discover and fetch documentation pages and links.

Key capabilities include:

• list_documentation_sources_tool: list all configured documentation sources (no arguments).
• fetch_documentation_page: fetch the content of a documentation page by URL as markdown (requires url).
• fetch_document_links: fetch all links from a documentation page (requires url).

Prompts to use these tools include: documentation_sources to list sources, documentation_page with a specific URL to retrieve content, and documentation_links to discover related content at a given URL.

Prerequisites:

• A supported runtime (uv/uvx) or Docker installed on your machine.
• Internet access to pull documentation sources.

Option A – Using uvx (recommended):

1. Ensure uv is installed and accessible, then install the mcp-server-docy package via uvx: uvx install mcp-server-docy
2. Run the server with the desired documentation URLs (example uses crawl4ai and React): DOCY_DOCUMENTATION_URLS="https://docs.crawl4ai.com/,https://react.dev/" uvx mcp-server-docy

Option B – Using Docker:

1. Pull the latest image: docker pull oborchers/mcp-server-docy:latest
2. Run the container with the documentation URLs set as an environment variable: docker run -i --rm -e DOCY_DOCUMENTATION_URLS="https://docs.crawl4ai.com/,https://react.dev/" oborchers/mcp-server-docy

Option C – Using Pip (alternative):

1. Install the package: pip install mcp-server-docy
2. Run the server as a module with the environment variable: DOCY_DOCUMENTATION_URLS="https://docs.crawl4ai.com/,https://react.dev/" python -m mcp_server_docy

Configuration tips:

• You can adjust the DOCY_CACHE_TTL and DOCY_CACHE_DIRECTORY to tune performance and storage.
• DOCY_HOST and DOCY_PORT let you bind the server to a specific interface and port if needed.

Tips and common considerations:

• Environment variables are the primary way to configure Docy (e.g., DOCY_DOCUMENTATION_URLS, DOCY_CACHE_TTL).
• If you need to update sources, edit the URLs file or re-run with a new DOCY_DOCUMENTATION_URLS value.
• For deployment, consider using the Docker option for isolated environments or the uvx option for quick local testing.
• If Claude defaults to WebFetchTool, add the recommended CLAUDE.md guidelines to explicitly request Docy functionality for documentation tasks.
• Supported transports are SSE or STDIO; adjust DOCY_TRANSPORT if you require a different data channel.
• Ensure URL sources are reachable from your deployment environment and comply with site scraping policies.