MCP Qdrant Server with OpenAI Embeddings

This MCP server provides vector search capabilities using Qdrant vector database and OpenAI embeddings.

Features

• Semantic search in Qdrant collections using OpenAI embeddings
• List available collections
• View collection information

Prerequisites

• Python 3.10+ installed
• Qdrant instance (local or remote)
• OpenAI API key

Installation

1. Clone this repository:

   git clone https://github.com/yourusername/mcp-qdrant-openai.git
   cd mcp-qdrant-openai
   

2. Install dependencies:

   pip install -r requirements.txt
   

Configuration

Set the following environment variables:

• OPENAI_API_KEY: Your OpenAI API key
• QDRANT_URL: URL to your Qdrant instance (default: “http://localhost:6333”)
• QDRANT_API_KEY: Your Qdrant API key (if applicable)

Usage

Run the server directly

python mcp_qdrant_server.py

Run with MCP CLI

mcp dev mcp_qdrant_server.py

Installing in Claude Desktop

mcp install mcp_qdrant_server.py --name "Qdrant-OpenAI"

Available Tools

query_collection

Search a Qdrant collection using semantic search with OpenAI embeddings.

• collection_name: Name of the Qdrant collection to search
• query_text: The search query in natural language
• limit: Maximum number of results to return (default: 5)
• model: OpenAI embedding model to use (default: text-embedding-3-small)

list_collections

List all available collections in the Qdrant database.

collection_info

Get information about a specific collection.

• collection_name: Name of the collection to get information about

Example Usage in Claude Desktop

Once installed in Claude Desktop, you can use the tools like this:

What collections are available in my Qdrant database?

Search for documents about climate change in my "documents" collection.

Show me information about the "articles" collection.