MCP server

Start the server

The included server communicates over stdio and exposes one tool named discover, giving your assistant a way to ask what exists in the world.

npm run mcp

Configure an MCP client

Use an absolute path to the API repository on your machine.

{
  "mcpServers": {
    "outerview": {
      "command": "node",
      "args": ["/absolute/path/to/outerview-api/mcp/server.js"],
      "env": {
        "OUTERVIEW_API_KEY": "YOUR_OUTERVIEW_API_KEY",
        "OUTERVIEW_API_BASE_URL": "https://api.outerview.ai"
      }
    }
  }
}

Call discover

The tool accepts the API fields plus api_key, local file paths in files, and timeout_ms. It returns the exact structured response as JSON text.

• Use area_id or h3_ids to constrain the search geographically.
• Use search_id to continue a previous query.
• Pass local CSV or GeoJSON paths in files; the server reads and uploads them.

{
  "query": "public gardens in Ottawa",
  "limit": 50,
  "answer": true
}