Graphiti-Memory

Graphiti-Memory MCP Server enables memory storage and knowledge graph queries against a Neo4j database, with optional AI-enhanced entity extraction. You can add memories, search for entities and relationships, retrieve episodes, and manage or clear the graph from any MCP-compliant client.

How to use

You interact with the Graphiti-Memory MCP Server from your MCP client. Start the server locally or connect to it via HTTP if you have a remote endpoint, then issue commands to store memories, search for entities, explore relationships, or retrieve historical episodes. Use the available tools to perform these actions and, if you enable it, leverage AI-powered extraction to enrich your graph with higher-quality entities and connections.

How to install

Prerequisites: you need a running Neo4j database and Python 3.10 or newer.

Install from PyPI

pip install graphiti-memory

Install from source

git clone https://github.com/alankyshum/graphiti-memory.git
cd graphiti-memory
pip install -e .

Neo4j prerequisites include setting up a password and ensuring the database is reachable on the default ports.

Configuration and operation

MCP configuration lets your client specify how to run the server and how to connect to Neo4j and optional AI services.

{
  "mcpServers": {
    "graphiti-memory": {
      "command": "graphiti-mcp-server",
      "env": {
        "NEO4J_URI": "neo4j://127.0.0.1:7687",
        "NEO4J_USER": "neo4j",
        "NEO4J_PASSWORD": "your-password-here",
        "OPENAI_API_KEY": "your-openai-key-here",
        "GRAPHITI_GROUP_ID": "default"
      }
    }
  }
}

Neo4j setup

Set up your Neo4j instance and ensure you can authenticate.

neo4j-admin dbms set-initial-password YOUR_PASSWORD

Standalone testing

You can test the server from the command line directly. Set environment variables for Neo4j and run the server to accept input via stdin.

export NEO4J_URI="neo4j://127.0.0.1:7687"
export NEO4J_USER="neo4j"
export NEO4J_PASSWORD="your-password"

echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}' | graphiti-mcp-server

Available tools

add_memory

Add a memory or episode to the knowledge graph, including content type, source, and optional grouping.

search_memory_nodes

Query for entities within the knowledge graph using natural language and retrieve matching nodes.

search_memory_facts

Query for relationships between entities to discover connections and patterns.

get_episodes

Retrieve recent memory episodes for a specific group.

delete_episode

Remove an episode from the knowledge graph by UUID.

delete_entity_edge

Delete a specific relationship (edge) between entities by UUID.

get_entity_edge

Fetch a specific entity edge by UUID.

clear_graph

Destructively remove all data from the knowledge graph.