mcp-registry/hannesrudolph/mcp-ragdocs

    ==================
      
       /// MCP ///
      /// HAN ///
        
    ==================
        
    [server:online]
    [protocol:ready]

hannesrudolph/mcp-ragdocs

by hannesrudolph

MCP server that performs RAG-style documentation retrieval via vector search (Qdrant) to enrich LLM responses.

212
25
Open Source
GitHub

01

search_documentation

Search stored documentation using a natural-language query and return the most relevant excerpts.

02

list_sources

Return a complete list of all documentation sources currently indexed in the system.

03

extract_urls

Crawl a specified webpage, extract every hyperlink found, and optionally enqueue them for later indexing.

04

remove_documentation

Permanently remove specified documentation sources from the database by their URLs.

05

list_queue

List all URLs that are currently waiting in the documentation processing queue.

06

run_queue

Process and index every URL presently in the queue, continuing until the queue is empty or an unrecoverable error occurs.

07

clear_queue

Immediately remove all pending URLs from the processing queue, effectively resetting it.

Installation

1. Clone the repo
git clone https://github.com/hannesrudolph/mcp-ragdocs.git cd mcp-ragdocs
2. Install JavaScript/TypeScript dependencies
# with npm npm install # or with pnpm pnpm install
3. Configure environment
cp .env.example .env # then edit .env and provide at least the following keys OPENAI_API_KEY=<your-openai-key> # (optional) VECTOR_DB_URL=<postgres/pinecone/chroma url> EMBEDDING_MODEL=text-embedding-3-small
4. Build TypeScript → JavaScript
npm run build
5. Start the MCP server
npm start # production build # or hot-reload for development npm run dev
6. The server starts on http://localhost:3100 (default). Set MCP_SERVER_URL to this address when you register the server with your assistant platform.

Documentation

License: MIT License
Updated 7/30/2025