Toronto MCP Server

Toronto MCP Server: Toronto Open Data Tools

This project implements a Model Context Protocol (MCP) server for Toronto Open Data, deployable on Cloudflare Workers. It exposes a comprehensive set of tools for intelligently querying, analyzing, and retrieving datasets from Toronto's CKAN-powered open data portal, making them accessible to MCP-compatible clients like Claude Desktop, Cursor, and other AI assistants.

🚀 Live Server

Deployed at: https://toronto-mcp.s-a62.workers.dev

• SSE Endpoint: https://toronto-mcp.s-a62.workers.dev/sse (for Claude Desktop)
• MCP Endpoint: https://toronto-mcp.s-a62.workers.dev/mcp (for other clients)

What does it do?

• Provides a remote MCP server exposing tools for Toronto's Open Data via the CKAN API
• Intelligently discovers relevant datasets using advanced relevance scoring
• Analyzes data freshness patterns with comprehensive update frequency tracking
• Provides deep data structure insights including field analysis and schema information
• Enables natural language querying of Toronto's 500+ open datasets
• Supports comprehensive data analysis combining multiple analytical dimensions

🛠️ Features

Basic CKAN Tools

• list_datasets: List all available datasets
• search_datasets: Search datasets by keyword
• get_package: Retrieve complete metadata for a dataset
• get_first_datastore_resource_records: Get records from the first active resource
• get_resource_records: Get records from a specific resource by ID

🧠 Advanced Analysis Tools

• find_relevant_datasets: Intelligently find and rank datasets using relevance scoring (title, description, tags, organization)
• analyze_dataset_updates: Analyze update frequencies with categorization (daily, weekly, monthly, quarterly, annually, irregular)
• analyze_dataset_structure: Deep-dive into dataset structure with field definitions, data types, record counts, and optional data previews
• get_data_categories: Explore all available organizations and topic groups
• get_dataset_insights: Comprehensive analysis combining relevance ranking, update frequency, and data structure insights

💡 Use Cases

For AI Assistants & Researchers

• "What traffic data is available in Toronto?" → Ranked datasets with relevance scores and update frequencies
• "How current is Toronto's environmental data?" → Update frequency analysis across environmental datasets
• "What fields are in the building permits dataset?" → Complete schema analysis with data types and sample records
• "Give me insights about Toronto's budget data" → Comprehensive analysis with relevance, freshness, and structure
• "Which datasets update daily?" → Frequency-based filtering and categorization

For Data Scientists & Analysts

• Discover datasets relevant to specific research questions
• Assess data quality and reliability through update patterns
• Understand data structure before detailed analysis
• Find related datasets across different city departments
• Evaluate data completeness and field availability

🏗️ Tech Stack

• Cloudflare Workers: Serverless deployment platform
• Model Context Protocol (MCP): Standard for AI tool integrations
• TypeScript: Type safety and modern development
• Zod: Runtime parameter validation
• CKAN API: Direct integration with Toronto Open Data

📁 Project Structure

toronto-mcp/
├── src/
│   ├── index.ts                 # MCP server setup and routing
│   └── ckanTools.ts            # Toronto Open Data tools implementation
├── test-runner.ts              # Automated testing framework
├── test-deployment.ts          # Deployment validation script
├── claude-mcp-config.json      # Claude Desktop configuration
├── evaluation-guide.md         # Comprehensive testing strategies
├── example-usage.md            # Usage examples and patterns
├── testing-guide.md            # Automated testing documentation
└── README.md                   # This file

🚀 Quick Start

1. Deploy Your Own Instance

# Clone and deploy
git clone <your-repo>
cd toronto-mcp
npm install
wrangler deploy

2. Test Deployment

# Install testing dependencies
npm install tsx

# Test your deployment
npx tsx test-deployment.ts https://your-worker.workers.dev

3. Connect to Claude Desktop

Create or edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "toronto-mcp": {
      "command": "npx",
      "args": ["mcp-remote", "https://toronto-mcp.s-a62.workers.dev/sse"]
    }
  }
}

Restart Claude Desktop and start asking questions about Toronto's open data!

🧪 Testing & Validation

Quick Connectivity Test

npx tsx test-deployment.ts https://toronto-mcp.s-a62.workers.dev

Automated Testing Framework

# Run mock tests (validates framework)
npm test

# Test specific deployment
npm run test