Home / MCP / Graph Memory MCP Server

Graph Memory MCP Server

A Neo4j-backed MCP server for storing and querying memory from AI interactions, with full MCP compatibility and Docker support.

typescript

20stars

View on GitHub

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
    "mcpServers": {
        "graph_memory": {
            "command": "npx",
            "args": [
                "-y",
                "@izumisy/mcp-neo4j-memory-server"
            ],
            "env": {
                "NEO4J_URI": "neo4j://localhost:7687",
                "NEO4J_USER": "neo4j",
                "NEO4J_PASSWORD": "password",
                "NEO4J_DATABASE": "neo4j"
            }
        }
    }
}

You have access to a Neo4j-backed MCP server that stores and retrieves memory from interactions with AI assistants. It delivers powerful graph querying, CRUD for entities and observations, and seamless MCP protocol compatibility. This guide shows you how to install, configure, and use the server in practical steps.

How to use

Connect your MCP client to the server using the provided command and environment settings. You will interact with a memory graph that stores entities, observations, and relationships, and you can run search queries that combine exact matches and fuzzy matching to retrieve relevant memory.

How to install

Prerequisites you need before installation are Node.js version 22.0.0 or newer and a running Neo4j database (local or remote). Ensure your environment can reach the Neo4j instance you plan to use.

# Global installation (as shown in the setup flow)
npm install -g @jovanhsu/mcp-neo4j-memory-server

# Or as a project dependency
npm install @jovanhsu/mcp-neo4j-memory-server

Additional configuration and deployment options

If you prefer a containerized deployment, you can use Docker to bring up Neo4j and the Memory Server together with a single command flow.

# Use docker-compose to start Neo4j and Memory Server together
git clone https://github.com/JovanHsu/mcp-neo4j-memory-server.git
cd mcp-neo4j-memory-server
docker-compose up -d

Environment variables

Configure the server with the following environment variables to point to your Neo4j instance and database.

NEO4J_URI=neo4j://localhost:7687
NEO4J_USER=neo4j
NEO4J_PASSWORD=password
NEO4J_DATABASE=neo4j

MCP connection through stdio (local)

You can run a local MCP server process and connect to it via the MCP protocol using a standard I/O interface. The command shown starts the server with the necessary environment for Neo4j access.

npx -y @izumisy/mcp-neo4j-memory-server
# Ensure the following environment is set in the running process or your shell/session:
# NEO4J_URI=neo4j://localhost:7687
# NEO4J_USER=neo4j
# NEO4J_PASSWORD=password
# NEO4J_DATABASE=memory

What you can do with the server

- Create, read, update, and delete entities, relationships, and observations in the knowledge graph.

- Perform complex graph queries and traversals to retrieve context across multiple related items.

Available tools

create_entities

Create entities in the knowledge graph with associated observations.