This fork of the original Graphiti MCP Server is tailored for containerized deployments via Coolify, Docker Compose, or similar platforms.

Graphiti is a framework for building and querying temporally-aware knowledge graphs, specifically tailored for AI agents operating in dynamic environments. Unlike traditional retrieval-augmented generation (RAG) methods, Graphiti continuously integrates user interactions, structured and unstructured enterprise data, and external information into a coherent, queryable graph.

This fork streamlines Graphiti's MCP server into a Docker-first setup optimized for deployment orchestration, including Neo4j integration and SSE-based MCP transport.

• Dockerfile tailored for uv-based Python builds
• Neo4j + Graphiti containerized setup via Docker Compose
• Preconfigured environment variable support with .env
• Out-of-the-box compatibility with Coolify
• Healthcheck-based service orchestration

• Docker + Docker Compose
• OpenAI API key

git clone https://github.com/tn-py/graphiti-mcpserver-coolify.git
cd graphiti-mcpserver-coolify

cp .env.example .env

Set your OpenAI key and model preferences inside .env:

OPENAI_API_KEY=your_openai_key
MODEL_NAME=gpt-4.1-mini

docker compose up --build

This will:

• Launch Neo4j (v5.26) with dev memory presets
• Launch Graphiti MCP server
• Bind Graphiti to port 8000 (mapped to host port 3010)

Access the MCP SSE endpoint:

http://localhost:3010/sse

.
├── Dockerfile                # Graphiti MCP build
├── docker-compose.yml       # Multi-service stack (Neo4j + Graphiti)
├── .env.example             # Env config template
├── graphiti_mcp_server.py   # Entry point
└── pyproject.toml           # Python deps

To deploy on Coolify:

1. Connect your GitHub repo containing this fork
2. Ensure docker-compose.yml is in the root
3. Set your environment variables via Coolify UI or .env
4. Deploy the app

Coolify will:

• Build the Graphiti service using the Dockerfile
• Launch Neo4j in a linked container
• Expose Graphiti on a public port (e.g., 3010)

Configure your MCP-compatible client like Cursor or Claude to point to:

{
  "mcpServers": {
    "graphiti-memory": {
      "transport": "sse",
      "url": "http://localhost:3010/sse"
    }
  }
}

All upstream functionality remains intact:

• add_episode
• search_nodes, search_facts
• get_episodes, delete_episode
• clear_graph, get_status

• Docker Engine
• Python 3.10+ (for local development)
• Neo4j 5.26+
• OpenAI API key

This fork retains the original license from the Graphiti project.

Forked from the excellent work by the team at Zep.