A collection of Model Context Protocol (MCP) servers that provide graph data science capabilities through Neo4j Graph Data Science (GDS). These servers enable AI agents to perform graph analytics workflows including loading graphs, running algorithms, and managing GDS sessions.
Direct graph loading and analysis server. Load graphs programmatically via JSON data structures or Supabase queries, then run graph algorithms.
Key Features:
- Load graphs directly from structured data (nodes/relationships as JSON)
- Fetch graph data from Supabase databases
- Run algorithms: PageRank, Betweenness, WCC, Leiden
- Session and graph management
Install: uvx mcp-neo4j-aura-gds-standalone
File-based graph loading server. Load graphs from JSON files or fetch data from Supabase and save to files for later analysis.
Key Features:
- Load graphs from JSON files
- Fetch and persist graph data from Supabase
- Run algorithms: PageRank, Betweenness, WCC, Leiden
- Session and graph management
Install: uvx mcp-neo4j-aura-gds-standalone-file
Both servers require Neo4j Aura API credentials. Add to your claude_desktop_config.json:
{
"mcpServers": {
"neo4j-gds": {
"command": "uvx",
"args": ["mcp-neo4j-aura-gds-standalone"],
"env": {
"CLIENT_ID": "<your-aura-client-id>",
"CLIENT_SECRET": "<your-aura-client-secret>",
"CLOUD_PROVIDER": "gcp",
"CLOUD_REGION": "europe-west1"
}
}
}
}Get Aura API credentials from Neo4j Aura Console:
- Navigate to account settings
- Create API credentials (Client ID and Client Secret)
- Optionally specify a Project ID to scope sessions
# Install uv
pip install uv
# Clone and setup
git clone https://github.com/neo4j-contrib/mcp-aura-gds-standalone.git
cd mcp-aura-gds-standalone
uv venv
source .venv/bin/activate
uv pip install -e ".[dev]"MIT License - see LICENSE file for details.