DataGraph MCP Server

Model Context Protocol server for accessing the DataGraph API from Claude Desktop, ChatGPT, and other MCP-compatible clients.

Installation

Via npx (recommended)

{
  "mcpServers": {
    "datagraph": {
      "command": "npx",
      "args": ["-y", "datagraph-city-mcp-server"],
      "env": {
        "DATAGRAPH_API_KEY": "dgc_your_api_key_here"
      }
    }
  }
}

Via local install

npm install datagraph-city-mcp-server

{
  "mcpServers": {
    "datagraph": {
      "command": "node",
      "args": ["/absolute/path/to/node_modules/datagraph-city-mcp-server/index.js"],
      "env": {
        "DATAGRAPH_API_KEY": "dgc_your_api_key_here"
      }
    }
  }
}

Get an API key

Sign up at datagraph.city to get your free API key.

Claude Desktop setup

1. Open Claude Desktop config:
   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
2. Add the configuration above with your API key
3. Restart Claude Desktop

Example queries

"What programs address violence prevention in Kansas City?"
"Show me the obstacles to reducing urban loneliness in NYC"
"Which organizations fund violence prevention in Kansas City?"
"What strategy areas do Kansas City nonprofits work in?"
"Show recent building permits in Manhattan"
"What are property sale prices by borough in NYC?"

Available datasets

Call list_datasets first — it returns exact dataset names and current counts.

NYC dataset details

GOSR Framework

Goal-Obstacles-Solutions-Resources (GOSR) is a participatory problem-structuring method for civic problem-solving. Learn more at gosr.ai.

Key relationship chain:

(Goal)-[:HAS_OBSTACLE]->(Obstacle)-[:HAS_SOLUTION]->(Solution)
    <-[:IMPLEMENTS]-(Resource)<-[:EXECUTES]-(Actor)<-[:FUNDS]-(Funder)
                    (Resource)<-[:FUNDS]-(Funder)  [when source names a specific program]
(Actor)-[:WORKS_IN]->(StrategyArea)

FUNDS edge targeting:

• (Funder)-[:FUNDS]->(Actor) — when the funding source names only the organization (most sources)
• (Funder)-[:FUNDS]->(Resource) — when the funding source names a specific program (currently COMBAT grants only)
• To find all funding, query both: MATCH (f)-[:FUNDS]->(n) WHERE n:Actor OR n:Resource

Critical distinctions:

• Solutions are theoretical/potential interventions — they describe what could work
• Resources are real, existing programs currently operating
• Obstacles can have child Solutions (leaf nodes) OR child Obstacles (sub-problems), but not both

Counting parent vs. leaf Obstacles:

// Parent obstacles (have sub-obstacles)
MATCH (p:Obstacle)-[:HAS_OBSTACLE]->(c:Obstacle)
WHERE p.dataset = 'your-dataset'
RETURN COUNT(DISTINCT p)

// Leaf obstacles (have solutions)
MATCH (p:Obstacle)-[:HAS_OBSTACLE]->(c:Obstacle)
WHERE p.dataset = 'your-dataset'
RETURN COUNT(c)

Available tools

query_locality_data parameters

Tip: Call get_locality_schema first to understand the graph structure, then write Cypher directly for precise results.

Development

Test with MCP Inspector

npx @modelcontextprotocol/inspector node index.js

Debug mode

DEBUG=* node index.js

Troubleshooting

"API key not found" — Ensure DATAGRAPH_API_KEY is set in your MCP server env config.

Claude doesn't see the server — Use absolute paths. Restart Claude Desktop completely after config changes.

Empty query results — Use get_locality_schema first to understand the graph structure, then write Cypher directly via query_locality_data.

"Command failed" — Verify Node.js is installed (node --version), then test manually: node index.js.

Support

• Documentation: docs.datagraph.city
• Issues: github.com/team-earth/datagraph-city-mcp-server/issues
• Email: support@datagraph.city