Qdrant MCP Server
1,540 starsQdrant's official MCP server: turns the Qdrant vector database into a semantic memory layer for agents — qdrant-store saves information, qdrant-find retrieves by semantic similarity. Connects to a Qdrant server (QDRANT_URL) or a local embedded database (QDRANT_LOCAL_PATH); embeddings are computed locally with fastembed (default sentence-transformers/all-MiniLM-L6-v2).
- First published：2024-12-02
- data as of：
- 2026-09-30
- Licence：
- Apache-2.0
- Transport：
- stdio
- Primitives exposed：
- tools
What it can touch
Reads and writes the Qdrant collection you name (storing/retrieving memory entries); local mode reads and writes an embedded database file on this machine; server mode sends content to your Qdrant instance. First use downloads the embedding model from Hugging Face.
Risk
Same poisoning risk as any memory layer: false or malicious entries are semantically recalled again and again, shaping every later decision. The model download is a supply-chain point. In server mode, memory contents leave the machine for your Qdrant instance.
Install prompt
從官方倉庫 README 取得執行方式（uvx 或 Docker），在客戶端設定中以環境變數傳入 QDRANT_URL（或 QDRANT_LOCAL_PATH）與 COLLECTION_NAME，然後重啟客戶端。首次執行會下載內嵌模型。Paste this to your agent. It deliberately contains no tool subcommands — check the tool's official docs for commands.