mcp-neo4j-memory-claude

A knowledge graph memory implementation for Claude AI using Neo4j and the Model Context Protocol (MCP).

This package provides a persistent memory store that allows Claude to save, retrieve, and reason about structured knowledge in conversations. It's based on the official Neo4j MCP implementation.

Features

• Persistent Memory: Store conversation knowledge across sessions
• Knowledge Graph Structure: Entity-relationship model for structured data
• Semantic Search: Find relevant information and connections
• Contextual Memory: Add observations to existing entities
• Relationship Tracing: Track how concepts connect to each other

Installation

npm install -g mcp-neo4j-memory-claude

Prerequisites

• Neo4j database instance (cloud or local)
• Claude AI with MCP support (Claude 3 Opus/Sonnet via Claude Desktop)

Configuration

Environment Variables

You'll need to provide your Neo4j credentials:

• NEO4J_URI: Connection URI for your Neo4j instance
• NEO4J_USER: Username for Neo4j database
• NEO4J_PASSWORD: Password for Neo4j database

Claude Desktop Integration

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "mcp-neo4j-memory": {
      "command": "npx",
      "args": [
        "mcp-neo4j-memory-claude"
      ],
      "env": {
        "NEO4J_URI": "neo4j+s://your-instance-id.databases.neo4j.io",
        "NEO4J_USER": "your_username",
        "NEO4J_PASSWORD": "your_password"
      }
    }
  }
}

Available Tools

This MCP server provides the following tools for Claude:

| Tool | Description | |------|-------------| | create_entities | Create new entities in the knowledge graph with observations | | create_relations | Create relationships between existing entities | | add_observations | Add new observations to existing entities | | delete_entities | Remove entities and their relationships | | delete_observations | Remove specific observations from entities | | delete_relations | Remove relationships between entities | | read_graph | Retrieve the entire knowledge graph | | search_nodes | Find entities matching search criteria | | open_nodes | Retrieve specific entities by name |

Example Usage (in Claude)

I'd like to build a knowledge graph about science fiction authors.

@mcp-neo4j-memory create_entities
[
  {
    "name": "Isaac Asimov",
    "entityType": "Author",
    "observations": ["Born in 1920", "Known for Foundation series", "Wrote about the Three Laws of Robotics"]
  },
  {
    "name": "Foundation",
    "entityType": "Book Series",
    "observations": ["Epic science fiction series", "Spans centuries of future history"]
  }
]

@mcp-neo4j-memory create_relations
[
  {
    "from": "Isaac Asimov",
    "to": "Foundation",
    "relationType": "WROTE"
  }
]

License

MIT

Contributing

Contributions welcome! Please check the GitHub repository for guidelines.

Acknowledgments

Based on the work by the Neo4j and Anthropic teams to integrate graph databases with Large Language Models through the Model Context Protocol.