neo4j-contrib/gds-agents
3.3
If you are the rightful owner of gds-agents and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.
This MCP Server leverages the Neo4j Graph Data Science (GDS) library to enable complex graph analysis and algorithm execution on Neo4j databases.
Tools
neo4j-gds:find_shortest_path
Finds the shortest path between two nodes in the graph.
neo4j-gds:degree_centrality
Calculates the degree centrality for nodes in the graph.
neo4j-gds:betweenness_centrality
Calculates the betweenness centrality for nodes in the graph.
neo4j-gds:closeness_centrality
Calculates the closeness centrality for nodes in the graph.
neo4j-gds:pagerank
Calculates the PageRank for nodes in the graph.
gds-agents
Neither LLMs nor any existing toolings (MCP Servers) are capable of complex reasoning on graphs at the moment. By reasoning on graphs, we mean the ability to decide and accurate execute the appropriate parameterised graph algorithms over large heterogeneous attributed graphs.
This MCP Server includes toolings from Neo4j Graph Data Science (GDS) library, which allows you to run all common graph algorithms.
Once the server is running, you are able to ask any graph questions about your Neo4j graph and get answers.
Setup
- Install necessary packages in your python environment with
pip install -r requirements.txt - Install the Neo4j database with GDS plugin: Download the Neo4j Desktop from Neo4j Download Center Install the GDS plugin from the Neo4j Desktop Create a new database and start it
- Populate .env file with necessary credentials:
NEO4J_URI=bolt://localhost:7687 # or other database URL if running in Aura NEO4J_USERNAME=neo4j # or other customer name NEO4J_PASSWORD=your_password - Load the London Underground dataset with the following command:
python import_data.py
Connect to your DB and querying the graph from Neo4j workspace,
you should see:
Start the server
Run uv sync and run claude from command line when inside the mcp_server.
Example usage
An example usage where we can ask Claude (that have access to the gds-agent MCP server) questions regarding my graph. It finds and executes the algorithms with appropriate parameters. Intermediate outputs are also shown.
> My database contains a graph about the London underground. I have some questions about my data. Use the available tools. If you cannot decide which tool to use, ask for confirmation.
āŗ I can help you analyze your London underground graph data! I have access to several Neo4j GDS tools for graph analysis including:
- Node operations: Count nodes, get node properties
- Centrality algorithms: PageRank, degree centrality, betweenness centrality, closeness centrality, eigenvector centrality, harmonic centrality, ArticleRank, HITS
- Path algorithms: Find shortest paths between nodes
- Graph structure: Find articulation points, bridges
- Influence maximization: CELF algorithm
What specific questions do you have about your London underground data? For example:
- Which stations are most central/important?
- What are the shortest paths between stations?
- Which stations are critical connection points?
- How many nodes/stations are in your graph?
Let me know what you'd like to explore and I'll use the appropriate tools to analyze your data.
> How can I go from Embankment to Paddington quickly?
āŗ neo4j-gds:find_shortest_path (MCP)(start_node: "Embankment", end_node: "Paddington")ā¦
āæ {'totalCost': 6.0, 'nodeIds': [71, 251, 89, 25, 9, 68, 167], 'nodeNames': [<Node element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:71' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5074, 'rail': 0, 'name': 'Embankment', 'total_lines': 4, 'id': '87',
'display_name': 'Embankment', 'longitude': -0.1223}>, <Node element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:251' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.501, 'rail': 0, 'name': 'Westminster', 'total_lines': 3, 'id': '285', 'display_name': 'Westminster',
'longitude': -0.1254}>, <Node element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:89' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5067, 'rail': 0, 'name': 'Green Park', 'total_lines': 3, 'id': '107', 'display_name': 'Green<br />Park', 'longitude': -0.1428}>, <Node
element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:25' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5142, 'rail': 0, 'name': 'Bond Street', 'total_lines': 2, 'id': '28', 'display_name': 'Bond<br />Street', 'longitude': -0.1494}>, <Node
element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:9' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5226, 'rail': 0, 'name': 'Baker Street', 'total_lines': 5, 'id': '11', 'display_name': 'Baker<br />Street', 'longitude': -0.1571}>, <Node
element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:68' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5203, 'rail': 0, 'name': 'Edgware Road (C)', 'total_lines': 3, 'id': '83', 'display_name': 'Edgware<br />Road', 'longitude': -0.17}>, <Node
element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:167' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5154, 'rail': 1, 'name': 'Paddington', 'total_lines': 4, 'id': '193', 'display_name': 'Paddington', 'longitude': -0.1755}>], 'path': <Path start=<Node
element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:71' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5074, 'rail': 0, 'name': 'Embankment', 'total_lines': 4, 'id': '87', 'display_name': 'Embankment', 'longitude': -0.1223}> end=<Node
element_id='4:4da44164-1c21-4122-8b25-791a7359cb53:167' labels=frozenset({'UndergroundStation'}) properties={'zone': 1, 'latitude': 51.5154, 'rail': 1, 'name': 'Paddington', 'total_lines': 4, 'id': '193', 'display_name': 'Paddington', 'longitude': -0.1755}> size=6>, 'costs': [0.0, 1.0, 2.0, 3.0,
4.0, 5.0, 6.0]}
āŗ The quickest route from Embankment to Paddington takes 6 stops:
Embankment ā Westminster ā Green Park ā Bond Street ā Baker Street ā Edgware Road ā Paddington
Total cost: 6 stops
> How central is Piccadilly Circus in the London underground network?
āŗ neo4j-gds:degree_centrality (MCP)(nodes: ["Piccadilly Circus"])ā¦
āŗ neo4j-gds:betweenness_centrality (MCP)(nodes: ["Piccadilly Circus"])ā¦
āŗ neo4j-gds:closeness_centrality (MCP)(nodes: ["Piccadilly Circus"])ā¦
āŗ neo4j-gds:pagerank (MCP)(nodes: ["Piccadilly Circus"])ā¦
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā®
ā Tool use ā
ā ā
ā neo4j-gds:degree_centrality(nodes: ["Piccadilly Circus"]) (MCP) ā
ā Calculate degree centrality for all nodes in the graph ā
ā ā
ā Do you want to proceed? ā
ā āÆ 1. Yes ā
ā 2. Yes, and don't ask again for neo4j-gds:degree_centrality commands in /Users/sbr/gds-agents/mcp_server ā
ā 3. No, and tell Claude what to do differently (esc) ā
ā ā
ā°āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāÆ
...