Graphiti MCP Server
by getzep
The Graphiti MCP server exposes Graphiti's key functionality through the Model Context Protocol (MCP), allowing AI assistants to interact with Graphiti's knowledge graph capabilities. It's an experimental implementation designed for AI agents operating in dynamic environments.
Last updated: N/A
What is Graphiti MCP Server?
The Graphiti MCP server is an experimental implementation that exposes Graphiti's key functionality through the Model Context Protocol (MCP). It enables AI assistants to interact with Graphiti's knowledge graph, allowing them to manage episodes, entities, search for facts, manage groups, and maintain the graph.
How to use Graphiti MCP Server?
To use the Graphiti MCP server, you need to clone the Graphiti repository, install the prerequisites (Python 3.10+, Neo4j, OpenAI API key), configure the server using environment variables, and then run the server using uv or Docker Compose. Finally, configure your MCP-compatible client (e.g., Claude, Cursor) to connect to the server using either stdio or SSE transport.
Key features of Graphiti MCP Server
Episode Management (add, retrieve, delete)
Entity Management (search, manage nodes and relationships)
Search Capabilities (semantic and hybrid search for facts and node summaries)
Group Management (organize data with group_id filtering)
Graph Maintenance (clear graph and rebuild indices)
Use cases of Graphiti MCP Server
Interactive, context-aware AI applications
AI assistants with persistent memory
Knowledge graph querying for AI agents
Integrating user interactions and enterprise data into a coherent graph
FAQ from Graphiti MCP Server
What are the prerequisites for running the Graphiti MCP server?
What are the prerequisites for running the Graphiti MCP server?
Python 3.10 or higher, a running Neo4j database (version 5.26 or later), and an OpenAI API key.
How do I configure the server?
How do I configure the server?
The server is configured using environment variables such as NEO4J_URI, NEO4J_USER, NEO4J_PASSWORD, and OPENAI_API_KEY. These can be set in a .env file or directly in your environment.
How do I run the server?
How do I run the server?
You can run the server directly using uv run graphiti_mcp_server.py or deploy it using Docker Compose.
How do I integrate the server with an MCP client?
How do I integrate the server with an MCP client?
Configure your MCP client to connect to the server using either stdio or SSE transport, providing the appropriate command or URL.
What tools are available through the MCP server?
What tools are available through the MCP server?
The server exposes tools such as add_episode, search_nodes, search_facts, delete_entity_edge, delete_episode, get_entity_edge, get_episodes, clear_graph, and get_status.