What features does Qdrant MCP server provide?

Q: What features does Qdrant MCP server provide?

qdrant-store tool: Stores information (string) with optional JSON metadata into a specified Qdrant collection and returns a confirmation message.. qdrant-find tool: Retrieves relevant information from Qdrant using a query and a collection, returning results as separate messages.. Semantic memory layer on top of Qdrant: Provides memory storage and retrieval capabilities, acting as a semantic memory layer for memories stored in Qdrant.. Automatic collection creation: Automatically creates the specified collection in Qdrant if it does not already exist.. Embedding model configuration: Configurable embedding provider (default fastembed) and embedding model (default sentence-transformers/all-MiniLM-L6-v2); currently supports only FastEmbed models.. Transport protocol support: Supports stdio (default), SSE, and streamable-http transports for MCP communication.. Environment-based configuration: Configured via environment variables (e.g., QDRANT_URL, QDRANT_LOCAL_PATH, COLLECTION_NAME, EMBEDDING_MODEL); CLI arguments are deprecated.. FastMCP tooling and inspector: Based on FastMCP with developer tooling, including an MCP inspector for testing and debugging MCP servers.

Qdrant MCP Server 2025: Features & Setup Guide

Overview

This repository provides mcp-server-qdrant, an MCP server implementing a semantic memory layer on top of the Qdrant vector search engine. It keeps and retrieves memories in Qdrant via two MCP tools: qdrant-store and qdrant-find. The qdrant-store tool accepts information (string), optional metadata (JSON), and a collection_name; it stores the data and returns a confirmation message. The qdrant-find tool accepts a query string and a collection_name and returns relevant stored memories as separate messages. The server is configured via environment variables (for Qdrant endpoint, API key, default collection, embedding provider/model, etc.). It defaults to the fastembed embedding provider with the sentence-transformers/all-MiniLM-L6-v2 model and currently supports only FastEmbed models. It supports multiple transport protocols (stdio, SSE, streamable-http), with stdio as default, and CLI arguments are not supported anymore. It can automatically create a collection if it does not exist. The server can be used with clients like Cursor, Claude Code, and VS Code, and can be run locally or in Docker; development tools include an MCP inspector.

Features

qdrant-store tool

Stores information (string) with optional JSON metadata into a specified Qdrant collection and returns a confirmation message.

qdrant-find tool

Retrieves relevant information from Qdrant using a query and a collection, returning results as separate messages.

Semantic memory layer on top of Qdrant

Provides memory storage and retrieval capabilities, acting as a semantic memory layer for memories stored in Qdrant.

Automatic collection creation

Automatically creates the specified collection in Qdrant if it does not already exist.

Embedding model configuration

Configurable embedding provider (default fastembed) and embedding model (default sentence-transformers/all-MiniLM-L6-v2); currently supports only FastEmbed models.

Transport protocol support

Supports stdio (default), SSE, and streamable-http transports for MCP communication.

Environment-based configuration

Configured via environment variables (e.g., QDRANT_URL, QDRANT_LOCAL_PATH, COLLECTION_NAME, EMBEDDING_MODEL); CLI arguments are deprecated.

FastMCP tooling and inspector

Based on FastMCP with developer tooling, including an MCP inspector for testing and debugging MCP servers.