jestelle/sf-parking-mcp

SF Parking API

A REST API that provides access to San Francisco parking data via the SFMTA ArcGIS REST API.

🎉 Live API: https://sf-parking.vercel.app/api

Features

This server exposes three tools for querying SF parking blockface data:

• get_parking_by_bbox: Query parking data within a bounding box (lat/lon coordinates)
• get_parking_by_street: Search for parking by street name
• get_parking_by_location: Find parking near a specific point

Data includes street parking availability, rates, schedules, and location information for all SF parking zones.

Installation

Option 1: Using uv (Recommended)

# Install uv if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh

# Clone or download this repository
git clone <your-repo-url>
cd sf-parking

# Install dependencies
uv sync

Option 2: Using pip

pip install -e .

Option 3: Install from PyPI (after publishing)

pip install sf-parking-mcp

Usage

With Claude Desktop

Add this to your Claude Desktop configuration file:

MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "sf-parking": {
      "command": "uv",
      "args": [
        "--directory",
        "/ABSOLUTE/PATH/TO/sf-parking",
        "run",
        "server.py"
      ]
    }
  }
}

Or if installed via pip:

{
  "mcpServers": {
    "sf-parking": {
      "command": "python",
      "args": ["-m", "server"]
    }
  }
}

Standalone Testing

# Run the server directly
uv run server.py

Example Queries

Once connected, you can ask Claude questions like:

• "What's the parking availability on Market Street?"
• "Find parking near coordinates 37.7749, -122.4194"
• "Show me all parking in the area bounded by 37.77, -122.42 and 37.78, -122.41"

API Details

This server queries the SFMTA ArcGIS REST API:

• Endpoint: sfpark_ODS.BLOCKFACE_RATES_VW
• Data: Real-time parking blockface rates, availability, and location info
• Max records per query: 1000

Available Fields

• Location: Street name, address range, lat/lon, orientation
• Availability: Threshold values, availability messages
• Pricing: Rate details, schedules, thresholds
• Identifiers: Blockface ID, geometry

Publishing to PyPI (Free Hosting)

To make this server easily installable anywhere:

1. Create a PyPI account at https://pypi.org/account/register/
2. Install build tools: uv pip install build twine
3. Build the package: python -m build
4. Upload to PyPI: twine upload dist/*

After publishing, anyone can install with: pip install sf-parking-mcp

Development

# Install dev dependencies
uv sync

# Run server locally
uv run server.py

# Test with MCP inspector
npx @modelcontextprotocol/inspector uv run server.py

License

MIT

Contributing

Contributions welcome! Please open an issue or PR.