Qdrant MCP Server

Use it when

• Your agent forgets decisions, snippets, and notes between sessions and you want it to store and semantically retrieve them via MCP tools
• You want Cursor, Claude Code, or Claude Desktop to keep a searchable memory of reusable code snippets described in natural language

What it solves

Not the fit when

• Automatic capture of everything said between chats; the server exposes explicit `qdrant-store` and `qdrant-find` tools
• Standalone vector-database administration or bulk-ingestion pipelines
• Embedding providers other than FastEmbed models
• Write access when policy requires `QDRANT_READ_ONLY=true`
• Exposing SSE or streamable-HTTP transport to untrusted networks without a separate authentication and network-access boundary
• mapping or indexing an entire large codebase automatically
• improving embedding model recall quality
• hosting or scaling the Qdrant database itself
• document parsing or OCR before storage

Install

uvx mcp-server-qdrant (with QDRANT_URL or QDRANT_LOCAL_PATH plus COLLECTION_NAME env vars), or docker build -t mcp-server-qdrant . && docker run, or npx @smithery/cli install mcp-server-qdrant --client claude

Invoke

Add to your MCP client config (claude_desktop_config.json mcpServers, claude mcp add code-search -- uvx mcp-server-qdrant, or .vscode/mcp.json); the agent then calls the qdrant-store tool to save information and qdrant-find to retrieve it by natural language query

Alternatives

No reviewed alternatives recorded yet.