This MCP server provides a Neo4j-backed knowledge graph memory store that can be queried and updated via the MCP protocol. It stores entities, relationships, and observations in a graph, enabling fast graph traversals, fuzzy and exact searches, and memory-like CRUD operations suitable for AI assistants. You can connect to it using the MCP client and either run it directly with npx or via a containerized setup, then use the MCP tools to create, read, update, and delete memory entities and their relations. The server is designed to be MCP-compliant and integrates with tools like the MCP Inspector for debugging and exploration. Typical usage involves starting the server, configuring your MCP client to point at the server, and performing operations such as creating entities, establishing relationships, and querying memory for context.

In practice, you would run the server, then use the MCP client to call tools such as create_entities, get_memory, update_entity, and search_memory. The memory is stored in Neo4j, enabling complex graph queries and traversals over entities, relations, and observations. You can also leverage the provided environment variables to configure the Neo4j connection (URI, user, password, database) to suit local development or production deployments.

Prerequisites:

• Node.js >= 22.0.0
• Neo4j database (local or remote) running and accessible

Install steps (example):

1. Install globally or as a project dependency via npm: npm install -g @jovanhsu/mcp-neo4j-memory-server

   or as a project dependency:

   npm install @jovanhsu/mcp-neo4j-memory-server

2. Run the MCP server using npx (as shown in the README configuration): npx -y @jovanhsu/mcp-neo4j-memory-server

3. Verify the server is reachable (via MCP client or inspector) and ensure Neo4j is accessible with the configured credentials.

4. Optional: start with Docker/compose as described in the repository README if you prefer containerized deployment.

Environment variables:

• NEO4J_URI: URI for connecting to Neo4j (default bolt://localhost:7687)
• NEO4J_USER: Neo4j username (default neo4j)
• NEO4J_PASSWORD: Neo4j password (default password)
• NEO4J_DATABASE: Neo4j database name (default neo4j)

Tips and common issues:

• Ensure Neo4j is up and the credentials match what you provide in env vars.
• When using MCP Inspector, start the Neo4j Memory Server first, then start the inspector to connect.
• If you encounter connectivity issues, verify network accessibility and Neo4j's listening address.
• Use the MCP protocol tools like create_entities, get_memory, and search_memory to manage and query the knowledge graph memory.
• Docker deployment is supported; use docker-compose to spin up Neo4j and the memory server for easier orchestration.