Ragdocs (Vector Documentation Search)

🚀 RAG Documentation MCP Server

An MCP server implementation that provides vector-based search tools for retrieving and processing documents, enabling AI assistants to enhance their responses with relevant document context.

🚀 Quick Start

The RAG Documentation MCP Server offers a set of tools for document search, retrieval, and processing. It can be used to enhance AI responses with relevant document context, build document-aware AI assistants, and more.

✨ Features

• Vector-based document search and retrieval
• Support for multiple document sources
• Semantic search functionality
• Automated document processing
• Real-time context enhancement for LLMs (Large Language Models)

💻 Usage Examples

search_documentation

Search through stored documentation using a natural language query. Returns matching excerpts and their context, sorted by relevance.

Input:

• query (string): The text to search for in the documentation. Can be a natural language query, a specific term, or a code snippet.
• limit (number, optional): The maximum number of results to return (1 - 20, default: 5). A larger limit provides more comprehensive results but may increase processing time.

list_sources

List all document sources currently stored in the system. Returns a comprehensive list of all indexed documents, including the source URL, title, and last updated time. Use this tool to understand the available document content for search or to verify if a specific source has been indexed.

extract_urls

Extract and analyze all URLs from a specified web page. This tool crawls the specified page, identifies all hyperlinks, and optionally adds them to the processing queue.

Input:

• url (string): The full URL of the web page to analyze (must include the protocol, e.g., https://). The page must be publicly accessible.
• add_to_queue (boolean, optional): If true, automatically adds the extracted URLs to the processing queue for subsequent indexing. Use with caution on large websites to avoid overloading the queue.

remove_documentation

Permanently remove a specific document source by its URL. This operation affects future search results.

Input:

• urls (array of strings): A list of URLs to remove from the database. Each URL must exactly match the URL used when the document was added.

list_queue

List all URLs currently waiting in the document processing queue. Displays the pending document sources that will be processed when run_queue is called. Use this tool to monitor the queue status, verify if URLs were added correctly, or check the processing backlog.

run_queue

Process and index all URLs in the document queue. Each URL will be processed sequentially, with error handling and retry logic provided. Progress updates are provided during processing. Long-running operations will continue until the queue is empty or an unrecoverable error occurs.

clear_queue

Remove all pending URLs from the document processing queue. Use this tool to restart, remove unwanted URLs, or cancel pending operations. This action is immediate and permanent - URLs will need to be added again if processing is required later.

📚 Documentation

Use Cases

The RAG Documentation tools are suitable for:

• Enhancing AI responses with relevant document context
• Building document-aware AI assistants
• Creating context-aware tools for developers
• Implementing semantic document search
• Enhancing existing knowledge bases

Configuration

Using with Claude Desktop

Add the following content to claude_desktop_config.json:

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

The following environment variables need to be set:

• OPENAI_API_KEY: The key for communicating with the OpenAI API.
• QDRANT_URL: The URL of the Qdrant vector database service.
• QDRANT_API_KEY: The API key required to access the Qdrant service.

📄 License

This project is licensed under the MIT License.

Acknowledgments

Thanks to Hannes RAG for laying the foundation for this project.