Graphiti MCP Server

What is Graphiti MCP Server?

The Graphiti MCP Server allows AI agents to interact with a knowledge graph for persistent memory, entity extraction, and relationship tracking using the Graphiti framework. It is designed for developers building agentic systems that need a structured, queryable memory layer.

How to use Graphiti MCP Server?

Install the graphiti CLI via pipx (recommended) or set up a local development environment with venv. Copy and edit the .env file with Neo4j credentials and an OpenAI API key, then use graphiti compose and graphiti up to start the Docker containers (Neo4j database and MCP server). The MCP server communicates over SSE or Stdio transport.

Key features of Graphiti MCP Server

• Exposes Graphiti functionality via MCP (SSE or Stdio transport).
• Tools for adding and searching episodes, nodes, and facts.
• Supports custom entity type definitions for tailored extraction.
• CLI for project initialization, entity management, and Docker control.
• Uses Docker Compose for easy deployment of Neo4j and server.
• Leverages uv for fast dependency management.

Use cases of Graphiti MCP Server

• Persistent memory for AI agents across conversation turns.
• Extracting entities and relationships from unstructured text.
• Building knowledge graphs that evolve over time with new data.
• Integrating a knowledge graph backend into MCP‑compatible AI clients.

FAQ from Graphiti MCP Server

What are the prerequisites for running the server?

Python 3.10 or higher, Docker & Docker Compose, and the uv package manager.

How does the MCP server communicate with clients?

It supports both SSE (Server‑Sent Events) and Stdio transport, configurable via MCP.

What database does the server use?

It uses Neo4j as the graph database, run as a Docker container managed by Docker Compose.

How do I configure the server?

Edit the .env file in the cloned repository to set Neo4j credentials, OpenAI API key, and optional model settings.

What is the graphiti CLI used for?

The CLI handles project initialization, entity management, generating Docker Compose files, and starting/stopping services.