RagDocs

You run RagDocs MCP Server to enable retrieval-augmented generation for documentation. It stores documents in a vector store and lets you add, search, list, and delete documents with semantic similarity, using Ollama or OpenAI embeddings and a Qdrant backend for vector storage.

How to use

You interact with RagDocs MCP Server through an MCP client. Start the server using your preferred runtime and then issue commands to add documents, perform semantic searches, list stored items, or remove documents. The server handles automatic chunking and embedding generation, stores vectors in Qdrant, and computes similarity to answer your queries.

How to install

npm install -g @mcpservers/ragdocs

After installation, start RagDocs in stdio mode by invoking the runtime with the provided MCP entry. The server relies on a Qdrant instance and an embedding provider (Ollama by default). Ensure you have a running Qdrant instance and the embedding service you prefer before starting RagDocs.

Configuration and startup

Configure RagDocs with the runtime command and environment variables shown below. The runtime command starts the RagDocs MCP server and passes the MCP package for execution. You can point RagDocs to a local Qdrant instance and choose Ollama or OpenAI embeddings.

{
  "mcpServers": {
    "ragdocs": {
      "command": "node",
      "args": ["@mcpservers/ragdocs"],
      "env": {
        "QDRANT_URL": "http://127.0.0.1:6333",
        "EMBEDDING_PROVIDER": "ollama"
      }
    }
  }
}

Start and test locally

1. Ensure a Qdrant instance is running locally or accessible remotely. 2) Ensure an embedding provider is available (Ollama locally by default). 3) Start the RagDocs MCP server using your MCP client or by loading the configuration above. 4) Use the client to add documents, run semantic searches, list documents, or delete entries.

Available tools

The server provides these end-user operations to manage and query documentation:

Add a document

Add a document to the system with its URL, content, and optional metadata. RagDocs will chunk the content, embed the chunks, and store them for semantic search.

Search documents

Query the stored documents using natural language. RagDocs returns results ranked by semantic similarity, with optional filters and score thresholds.

List documents

List all stored documents with pagination and grouping options so you can review what is stored and how it is organized.

Delete a document

Remove a document from the RAG system by its URL. This also removes associated embeddings and metadata.

Environment and prerequisites

Prerequisites include Node.js 16 or higher, a running Qdrant instance (local Docker or cloud with API key), and an embedding provider (Ollama by default or OpenAI with an API key). Ensure you configure the Qdrant URL and the embedding provider in the runtime environment.

Security and notes

Keep your OpenAI API keys and Qdrant credentials secure. Do not expose API keys in logs or shared configurations. Use environment isolation for different environments (dev, staging, prod) and restrict access to the MCP server endpoints.

Troubleshooting

If you encounter connection issues to Qdrant, verify QDRANT_URL is reachable and that the Qdrant service is running. If embeddings fail, check that the selected provider (ollama or openai) is available and properly configured. Review MCP client logs for start-up errors or missing environment variables.

Notes

Automatic text chunking and embedding generation are performed by RagDocs to enable efficient semantic search across large documents.

Examples of configuration and usage

You can adapt the provided configuration to your environment by replacing the QDRANT_URL and EMBEDDING_PROVIDER values. The server supports both local Ollama embeddings and remote OpenAI embeddings, depending on your subscription and needs.

Available tools

add_document

Add a document with a URL, content, and optional metadata so RagDocs can store and index it for semantic search.

search_documents

Query stored documents using natural language to retrieve semantically relevant results.

list_documents

List all stored documents with pagination and grouping options.

delete_document

Delete a document from the RAG system by its URL.