RAG Documentation

You can run a dedicated MCP server that enables semantic, vector-based access to documentation sources and augments AI responses with contextual information. This server processes documentation, stores embeddings, and exposes tools to search and manage sources so your AI assistants stay informed with up-to-date context.

How to use

You use this MCP server by connecting an MCP client to the server you run locally or remotely. Once connected, you can search across indexed documentation, view source details, and manage the processing queue for new sources. The primary actions you’ll perform are searching documentation with natural language queries, listing current sources, and handling new URLs to index.

How to install

Prerequisites: install Node.js and npm on your machine. You also need access to an OpenAI API key and a vector database endpoint (such as a Qdrant instance) with its API key.

Add the MCP server configuration to your MCP client. Use the following JSON snippet as the configuration block for the client’s MCP setup. This config starts the rag_docs MCP server using npx and passes the required environment variables.

{
  "mcpServers": {
    "rag_docs": {
      "command": "npx",
      "args": [
        "-y",
        "@hannesrudolph/mcp-ragdocs"
      ],
      "env": {
        "OPENAI_API_KEY": "",
        "QDRANT_URL": "",
        "QDRANT_API_KEY": ""
      }
    }
  }
}

Additional setup and run notes

Fill in the environment variable values with your actual keys and URLs. After configuring, start the MCP server through your MCP client or runtime as appropriate for your environment. The server exposes tools to search documentation, list sources, extract URLs, and manage the processing queue to keep your documentation context fresh.

Notes on usage and capabilities

This MCP server supports vector-based documentation search, multiple documentation sources, semantic search, automated documentation processing, and real-time context augmentation for language models. Use the search tool to retrieve relevant excerpts, use list_sources to verify indexed sources, and use the queue tooling to manage how new documentation is ingested.

Available tools

search_documentation

Search through stored documentation using natural language queries. Returns matching excerpts with context, ranked by relevance.

list_sources

List all documentation sources currently stored in the system. Returns details like source URLs, titles, and last update times.

extract_urls

Extract and analyze all URLs from a given web page. Can crawl a page and optionally add extracted URLs to the processing queue.

remove_documentation

Remove specific documentation sources by their URLs. Removal is permanent and affects future search results.

list_queue

List all URLs currently waiting in the documentation processing queue.

run_queue

Process and index all URLs in the documentation queue with progress updates.

clear_queue

Remove all pending URLs from the documentation processing queue.