mcp-server-qdrant

Why we included this project

Qdrant's official MCP server turns the vector database into a semantic memory layer for LLM tools. Two tools handle the work: one stores information with optional metadata, the other finds relevant pieces back through vector similarity, so a chat client or coding agent can recall past context without you building a retrieval pipeline from scratch. Because it speaks the Model Context Protocol, it connects directly to MCP clients like Claude Desktop and Cursor over stdio, SSE, or streamable HTTP. Configuration runs mostly through environment variables, with a Qdrant URL or a local path for storage plus an embedding model and a default collection name. Teams already running Qdrant get a maintained, vendor-supported way to hand their assistants semantic memory instead of owning a custom connector.