Skip to content
MCP server

Qdrant MCP

Full Qdrant API as 58 MCP tools, with plain-text semantic, BM25 and hybrid search

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI, self-hosted
Runs
Locally
Version
2.12.0
Listing updated
3 days ago

About Qdrant MCP

Qdrant MCP is an open-source MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.

Where the official Qdrant server exposes two tools, store and find, this one wraps the API one operation per tool: collection lifecycle, point CRUD, payload and vector editing, snapshots, and server observability. The author's site lists 33 of the 58 tools as read-only and describes the project as a portfolio piece, run by you rather than hosted.

Search covers dense, sparse and multivectors, RRF and DBSF hybrid fusion, nested prefetch with reranking, MMR diversity, grouped queries, recommend and discover. With the optional embeddings extra you can pass plain text: semantic search is embedded locally with fastembed using the model a collection declares, and BM25 keyword search needs no embeddings on Qdrant 1.20.

Two advisor tools stand out. qdrant_collection_doctor checks a collection's status, index, memory, aliases and text setup and answers each finding with the tool call that fixes it, and qdrant_search_quality_eval measures recall@k of approximate search against exact search. Safety settings include toolset selection, a read-only mode that removes every non-read-only tool, a cap on points per call, and an allowlist of origins for snapshot restore. The default toolset's schemas are about 48 KB.

Install

Run in a terminal

claude mcp add mcp-qdrant --env QDRANT_API_KEY='<QDRANT_API_KEY>' -- uvx mcp-qdrant

Add to claude_desktop_config.json

{
  "mcpServers": {
    "mcp-qdrant": {
      "command": "uvx",
      "args": [
        "mcp-qdrant"
      ],
      "env": {
        "QDRANT_API_KEY": "<QDRANT_API_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "mcp-qdrant": {
      "command": "uvx",
      "args": [
        "mcp-qdrant"
      ],
      "env": {
        "QDRANT_API_KEY": "<QDRANT_API_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "mcp-qdrant": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "mcp-qdrant"
      ],
      "env": {
        "QDRANT_API_KEY": "<QDRANT_API_KEY>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "mcp-qdrant": {
      "command": "uvx",
      "args": [
        "mcp-qdrant"
      ],
      "env": {
        "QDRANT_API_KEY": "<QDRANT_API_KEY>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "mcp-qdrant": {
      "command": "uvx",
      "args": [
        "mcp-qdrant"
      ],
      "env": {
        "QDRANT_API_KEY": "<QDRANT_API_KEY>"
      }
    }
  }
}

Configuration

  • QDRANT_URLQdrant server URL, e.g. http://localhost:6333 (mutually exclusive with QDRANT_LOCAL_PATH)
  • QDRANT_API_KEYQdrant API key, if your instance requires one
  • QDRANT_LOCAL_PATHPath to an embedded Qdrant instance, instead of a URL
  • QDRANT_MCP_TOOLSETSComma-separated toolsets to enable: core, search, payload, admin, snapshots, observability
  • QDRANT_MCP_READ_ONLYRemove every tool not marked read-only from the registry
  • QDRANT_MCP_MAX_LIMITLargest number of points one tool call may ask for (minimum 30); bigger requests are refused, not cut
  • QDRANT_MCP_SNAPSHOT_ALLOWED_ORIGINSComma-separated http(s) origins qdrant_snapshot_recover may restore from, besides file:// and QDRANT_URL's own origin
  • FASTEMBED_CACHE_PATHWhere text-embedding models are cached (semantic search needs the mcp-qdrant[embeddings] extra)

Tools this server provides (3)

Taken from the project documentation; the running server may differ.

qdrant_collection_doctor
Checks a collection's status, index, memory, aliases and text setup and suggests the fixing tool call.
qdrant_search_quality_eval
Measures recall@k of approximate versus exact search, plus latency.
qdrant_snapshot_recover
Restores a snapshot from file, the Qdrant URL's origin or allowed origins.

What it does

One tool per operation
Wraps the Qdrant API one-to-one rather than just store and find.
Plain-text search
Semantic, BM25 and hybrid queries from text.
Search advisor
Diagnose a collection and measure recall against exact search.
Toolset selection
Enable core, search, payload, admin, snapshots or observability groups.
Read-only mode
Remove every tool not marked read-only.
Request cap
Calls asking for more than the configured point limit are refused.
Snapshots
Per-collection and full-storage backup and restore.

Good for

  • Create and tune a Qdrant collection from chat
  • Run hybrid search over an existing collection
  • Diagnose why a collection's search seems slow
  • Back up and restore snapshots
  • Inspect a database safely in read-only mode

Pricing

Open source

Free; run it yourself

Prices change. Check mcp-qdrant.vazquezlabs.com before you buy.

Qdrant MCP FAQ

How is it installed?

From PyPI as mcp-qdrant, with Docker, or as a Claude Desktop bundle.

Does it need a hosted service?

No. The site says there is no hosted instance; you run it against your own Qdrant via QDRANT_URL or QDRANT_LOCAL_PATH.

Is it a RAG pipeline?

No. The README says it never parses documents or decides how to chunk text.

How do I limit what an agent can do?

Set QDRANT_MCP_READ_ONLY, pick toolsets with QDRANT_MCP_TOOLSETS, or lower QDRANT_MCP_MAX_LIMIT.

How is it tested?

The site reports 766 tests, with CI against a real Qdrant server.

Is Qdrant MCP free?

Yes. Qdrant MCP is open source, so you can use and self-host it for free.

How do I add Qdrant MCP to Claude Code?

Run this command in your terminal: claude mcp add mcp-qdrant --env QDRANT_API_KEY='<QDRANT_API_KEY>' -- uvx mcp-qdrant

What are the best alternatives to Qdrant MCP?

The closest alternatives to Qdrant MCP are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and Midplane.

Alternatives to Qdrant MCP

  1. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  2. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  3. MCP server for searching the Claude Code Ultimate Guide from any MCP client

    Runs locally
    • Open source
    • 6.5K GitHub stars
    • AI models and LLM tooling
  4. Local-first long-term memory layer for AI agents, served over MCP

    Not checked yet
    • 5 tools
    • Remote
    • Open source
    • 1.3K GitHub stars
    • Knowledge and memory
  5. Midplane

    MCP server

    Postgres gateway for AI agents with table grants, column masking and approval for writes

    Runs locally
    • Open source
    • 5 GitHub stars
    • Databases
  6. note1

    MCP server

    MCP server giving AI assistants search and summaries across your note1 meetings

    • 4 tools
    • Remote
    • Free trial
    • 1 GitHub star
    • Productivity