This project provides an official Qdrant implementation of the Model Context Protocol (MCP), enabling LLM applications to seamlessly integrate with Qdrant for semantic memory. It targets developers building AI-powered tools like IDEs or chat interfaces, offering a standardized way to connect LLMs with external data sources.

How It Works

The server acts as a semantic memory layer over Qdrant. It exposes two primary tools: qdrant-store for persisting information and metadata into a specified Qdrant collection, and qdrant-find for retrieving relevant information based on a query. It leverages fastembed for generating embeddings, with sentence-transformers/all-MiniLM-L6-v2 as the default model. Configuration is primarily managed via environment variables.

Quick Start & Requirements

• Install/Run: uvx mcp-server-qdrant
• Prerequisites: Qdrant instance (URL or local path), uvx (Python package manager).
• Configuration: Environment variables like QDRANT_URL, COLLECTION_NAME, EMBEDDING_MODEL.
• Docs: https://github.com/qdrant/mcp-server-qdrant

Highlighted Details

• Supports stdio (default) and sse (Server-Sent Events) transport protocols.
• Integrates with Claude Desktop, Cursor, Windsurf, and VS Code via MCP configurations.
• Can be configured as a specialized code search tool by customizing tool descriptions.
• Offers a development mode with an MCP inspector for testing and debugging.

Maintenance & Community

This is an official Qdrant project. Further community and contribution details are available on the GitHub repository.

Licensing & Compatibility

• License: Apache License 2.0.
• Compatibility: Permissive license allows commercial use and integration with closed-source applications.

Limitations & Caveats

Currently, only fastembed models are supported for embedding generation. The project notes that tool descriptions may require customization for specific use cases.