MCP Memgraph

What memgraph/mcp-memgraph MCP server does

The memgraph/mcp-memgraph MCP server connects an MCP-based application to Memgraph. Its primary database operation is run_query, which executes Cypher queries. This lets an agent retrieve graph data or perform other supported Cypher operations through the MCP interface.

The server also exposes schema inspection capabilities. Agents can search the graph schema with a regular expression, retrieve the complete schema for a node identified by its labels, inspect a relationship schema, and retrieve an enum definition by name. These operations give an agent a way to understand the graph structure before forming a query.

The project is part of Memgraph's AI Toolkit and is focused on Memgraph rather than providing a general-purpose database abstraction. The accompanying toolkit documentation identifies Memgraph as its primary database target.

How it works

The memgraph/mcp-memgraph MCP server is distributed as both a Python package named mcp-memgraph and a Docker image named memgraph/mcp-memgraph. The README shows two Docker transport modes. HTTP mode publishes the service on port 8000, while stdio mode keeps the process interactive and sets MCP_TRANSPORT=stdio for MCP clients that launch local processes.

In an MCP conversation, a model can call run_query for Cypher execution or use one of the schema tools to obtain structural information. The available operations are:

• run_query: execute a Cypher query.
• search_schema: find schema information using a regular expression.
• get_node_schema: retrieve a node schema by labels.
• get_relationship_schema: retrieve a relationship schema.
• get_enum_schema: retrieve an enum schema by name.

The supplied material does not describe the server's request payloads, response formats, connection settings, or database credential handling. Those details should be verified in the linked MCP documentation before deployment.

Setup and configuration

The documented Python installation is:

pip install mcp-memgraph

For a containerized HTTP deployment, the README provides:

docker run --rm -p 8000:8000 memgraph/mcp-memgraph:latest

For a containerized stdio deployment, it provides:

docker run --rm -i -e MCP_TRANSPORT=stdio memgraph/mcp-memgraph:latest

The same README uses a separate Memgraph container for local development, exposing Bolt on port 7687 and enabling schema information with --schema-info-enabled=true. It does not explicitly document how that database endpoint is passed to the MCP server, so the connection configuration cannot be inferred from the supplied excerpt.

Limitations and notes

The provided material lists Memgraph as the supported database target. MySQL and PostgreSQL appear only as source databases for the separate sql2graph migration agent, not as databases supported by this MCP server.

The toolkit FAQ says an LLM API key is needed for features such as entity extraction and natural-language queries in other toolkit components. It does not state that mcp-memgraph itself requires an LLM API key. No authentication method, license, version requirement, MCP client compatibility list, or production security guidance is specified in the excerpt.

The memgraph/mcp-memgraph MCP server is therefore a suitable fit when the required workflow is direct Memgraph access through MCP tools, especially when schema discovery is useful before query execution. Use the linked project documentation to confirm deployment-specific configuration and client setup.