Bay Wheels

This MCP Server provides real-time Bay Wheels bikeshare data through a modular MCP interface. You can use it to find the nearest bike or dock with available spaces, and it supports dockless bike checks as well. It exposes endpoints and a local command you can run for development and testing, and it can be deployed in containers or cloud platforms for production use.

How to use

Connect an MCP client to the Bay Wheels MCP Server to access live data. You can query for the nearest bike (classic or electric) and the nearest dock with available spaces. If you’re using a mobile client, point it to the MCP endpoint to start making requests such as locating nearby bikes or returning a bike near a chosen area.

Two primary ways exist to connect and run the server during development and testing: a local stdio server you run from your command line, and a configured MCP client that connects via HTTP to the deployed server.

For remote testing and production deployment, you can connect through the standard MCP HTTP URL path and use the dedicated MCP endpoint for transport. When you deploy, your client should target the endpoint at /mcp on the deployed server.

How to install

Prerequisites before you begin:

• Python (for running the MCP server) or a compatible runtime
• Access to a shell with permissions to run the server
• Docker and Docker Compose (optional for containerized deployment)

1. Set up the local MCP server command for development and testing.

{
  "mcpServers": {
    "bay_wheels": {
      "command": "/opt/homebrew/bin/uv",
      "args": [
        "--directory",
        "/path/to/bay-wheels-mcp",
        "run",
        "server.py"
      ]
    }
  }
}

2. Run the local MCP server directly for testing.

uv run server.py

Configuration and endpoints

Connections for clients are exposed via two primary methods. The HTTP method targets a deployed MCP endpoint, and the local stdio method lets you test from your development environment.

HTTP endpoint (for remote/mobile clients) ensure your client connects to the MCP path under the deployed URL.

Testing and health checks

Health checks help you verify that the server is running and reachable. When deployed via Docker, you can test the health endpoint at the default port.

# Test the health endpoint in a containerized or deployed environment
curl http://localhost:8000/health
# Example response: {"status":"healthy","service":"bay-wheels-mcp","version":"0.1.0"}

Tools and endpoints

The MCP server exposes specific endpoints to locate bikes and docks near a given location. The available tools are described in your MCP client as the following functions.

Connecting clients

To connect a client, configure it to point to the MCP URL with the correct transport. For mobile apps, use the endpoint path /mcp on your deployed server.

Available tools

find_nearest_bike

Finds the nearest bike availability given a location and optional bike type. Returns the closest bike data for quick pickup.

find_nearest_dock_spaces

Finds the nearest dock with available return spaces near a given location. Returns the closest matching docks and space counts.