graphiti-mcp-server

MCP.Pizza Chef: gifflet

Assistants normally start fresh every conversation. This gives one a memory that persists: facts you mention get stored as connected pieces rather than a flat pile, so later questions can draw on what came before and on how things relate. It runs against a graph database you host, works in Cursor, and needs an OpenAI key for the part that pulls facts out of text. Setup involves Docker, so it is a developer job.

Unmaintained · No commits in 14 months.

Data

Notes

Web/Research

Use This MCP server To

Have my assistant remember facts between chats Ask what it already knows about a project Build up notes on people and companies over time Connect related facts instead of storing a list Search everything my assistant has learned

README

Graphiti MCP Server 🧠

🌟 A powerful knowledge graph server for AI agents, built with Neo4j and integrated with Model Context Protocol (MCP).

🚀 Features

🔄 Dynamic knowledge graph management with Neo4j
🤖 Seamless integration with OpenAI models
🔌 MCP (Model Context Protocol) support
🐳 Docker-ready deployment
🎯 Custom entity extraction capabilities
🔍 Advanced semantic search functionality

🛠️ Installation

Prerequisites

Docker and Docker Compose
Python 3.10 or higher
OpenAI API key

Quick Start 🚀

Clone the repository:

git clone https://github.com/gifflet/graphiti-mcp-server.git
cd graphiti-mcp-server

Set up environment variables:

cp .env.sample .env

Edit .env with your configuration:

# Required for LLM operations
OPENAI_API_KEY=your_openai_api_key_here
MODEL_NAME=gpt-4o

Start the services:

docker compose up

graphiti-mcp-server FAQ

Can I use this to make my assistant remember things?

Yes — facts persist between conversations rather than resetting each session.

Do I need an account or key?

Yes, an OpenAI key is required for the part that extracts facts from what you write.

Where is my information stored?

In a Neo4j graph database you run yourself, so it stays on infrastructure you control.

Which apps does it work with?

It is documented for Cursor and works with other assistants supporting this connection.

How hard is it to set up?

You need Docker and a recent Python, so this is a developer setup rather than a quick install.