Memgraph MCP Server Is Now on Docker Hub

When we introduced the Memgraph MCP Server, the goal was straightforward: give LLMs a practical way to query and understand graph data in Memgraph. Since then, the server has grown from 2 tools to 12 tools, adding graph algorithms, vector search, and an experimental mode with elicitation and sampling.

The difference now is that trying it is much easier. You no longer need to clone the repo or set up Python just to get started. The Memgraph MCP Server is available on Docker Hub as memgraph/mcp-memgraph.

docker run -d -p 7687:7687 memgraph/memgraph-mage:latest --schema-info-enabled=true
docker run --rm -p 8000:8000 memgraph/mcp-memgraph:latest

That starts Memgraph, launches the MCP server, and exposes it at http://localhost:8000/mcp/ for VS Code, Cursor, Claude Desktop, or your own agents. This is if you are running it locally.

The Docker image removes most of the setup friction (outside of Docker) without changing how the server works. It runs over streamable HTTP by default, can switch to stdio with MCP_TRANSPORT=stdio, stays read-only with MCP_READ_ONLY=true, can use the experimental server with MCP_SERVER=memgraph-experimental, and can be pinned to a versioned tag such as memgraph/mcp-memgraph:0.1.11.

Out of the box, the server can run Cypher, inspect schema and configuration, check indexes, constraints, triggers, and procedures, traverse neighborhoods, run PageRank and betweenness centrality, and search node vectors. In practice, that gives an MCP client real access to live graph structure instead of just text.

If you want the fastest path, pull the image, point your MCP client at the endpoint, and start querying your graph. For full setup details, examples, and client configuration, see the GitHub documentation. If you prefer a local Python install, the package is also available on PyPI.

This Docker release is a small change on the surface, but it makes the Memgraph MCP Server much easier to evaluate, demo, and deploy. If you have been following our earlier MCP posts, this is the most direct way to try the current version.