Qdrant MCP Server
@davidwynter
A simple MCP server to access Qdrant
Overview
What is Qdrant MCP Server?
A dual-protocol server that provides both FastAPI and FastMCP clients for accessing Qdrant knowledge graph operations. It is designed for developers who need to interact with Qdrant vector databases using either REST APIs or the Model Context Protocol.
How to use Qdrant MCP Server?
Install dependencies with Poetry after cloning the repository. Configure environment variables (QDRANT_URL, OPENAI_API_KEY, etc.). Run the server using python -m qdrant_mcpserver.main with an optional --server-type argument to choose between FastAPI (default: FastMCP). Both servers expose the same endpoints.
Key features of Qdrant MCP Server?
- Flexible server selection via CLI argument
- Shared configuration and Qdrant client
- Consistent endpoints across both protocols
- Poetry for dependency management
- Environment variable configuration
- Production-ready build and deployment instructions
Use cases of Qdrant MCP Server?
- Upserting knowledge graph nodes into Qdrant
- Performing semantic vector search across nodes
- Deleting nodes by their IDs
- Running health checks on the server
FAQ from Qdrant MCP Server
What dependencies are required?
The server requires a running Qdrant instance (URL and optional API key), an OpenAI API key for embedding generation, and Python with Poetry for installation.
How do I choose between FastAPI and FastMCP?
Use the --server-type argument when running main.py. Accepted values are fastapi or fastmcp; FastMCP is the default.
What ports are used by default?
FastAPI runs on port 8000 by default; FastMCP runs on port 8080. Both are configurable via the PORT and MCP_PORT environment variables.
Are the API endpoints the same for both servers?
Yes. Both the FastAPI and FastMCP servers expose identical endpoints: POST /nodes/upsert, POST /nodes/search, DELETE /nodes, and GET /health.
