Build MCP Challenge - Fitness Routes MCP Server

This project is an MCP server that turns fitness goals into real routes. Give it “burn 300 calories near Texas A&M” and it calls Google Maps to build a walkable loop and return AI-ready context.

Comment

MCP Server: Google Maps + Fitness Routes

Purpose

This MCP server turns queries on route-finding and fitness activities into live Google Maps calls and returns structured context for an AI assistant. It can:

  • turn a query to “to burn X calories” into a followable route
  • fetch walking directions from Google Maps
  • optionally find nearby places to use as waypoints

The goal is to make the AI’s answer location-aware and goal-aware (calories → distance → real route).

Data Sources / APIs

  • Google Maps Geocoding API
  • Google Maps Directions API
  • Google Maps Places API

These are used to resolve text locations to coordinates, get real route distance/duration, and optionally find nearby places.

Requirements

  • Python 3.10+
  • A Google Maps Platform API key with the above APIs enabled
  • Packages from requirements.txt

Environment Variables

Create a .env file in the project root (This file can be pasted by copying .env.example):

GOOGLE_MAPS_API_KEY=YOUR_KEY_HERE
DEFAULT_ORIGIN=Memorial Student Center, College Station, TX
DEFAULT_DESTINATION=Century Tree, College Station, TX

Notes:

  • Replace YOUR_KEY_HERE with an actual API key from Google Maps in Google Cloud Platform.
  • DEFAULT_ORIGIN is used when the query does not specify a start.
  • You can change these to locations relevant to you.

Server Entry / How to Query

This MCP server is meant to be called by an MCP-capable client (for example Claude Desktop) that sends a natural-language query. The server receives the query, detects the intent (fitness route, directions, nearby place), fetches external data from Google Maps, and returns a structured context object that the AI can use to answer the user.

Using with Claude Desktop

If you want Claude Desktop to call this MCP server, add a new file (or update) called claude_desktop_config.json if it doesn't exist to register the server, for example:

{
  "mcpServers": {
    "maps-routes": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\path\\to\\fitness-routes",
        "run",
        "main.py"
      ]
    }
  }
}

Notes:

  • Change "C:\\path\\to\\fitness-routes" to actual route to this folder

Installation and Run

python -m venv .venv
.\.venv\Scripts\activate  # For Windows
# source .venv/bin/activate  # For Linux/MacOS
pip install -r requirements.txt
python main.py

This starts the MCP server. When you go into Claude, click on "Search and Tools" and make sure the MCP server titled "map-routes" is fully enabled.

How It Works

  1. The server receives a query/tool call.
  2. It detects intent, for example: a fitness route request.
  3. It converts calories → approximate distance (documented assumption).
  4. It calls Google Maps (geocoding, directions, places) to build a real route.
  5. It returns a context package that an AI can use directly.

Example Prompts

  • Route to burn 300 calories.
  • Give me a walking route from the Memorial Student Center that burns 400 calories.
  • Plan a walking loop near the Memorial Student Center so I end where I start.
  • Find a nearby gym from Texas A&M, College Station.
  • I have 30 minutes to walk right now, starting at the Memorial Student Center.

More Prompts

  • Route to burn 500 calories from Texas A&M, College Station.
  • I need a 4 km walk starting near the Memorial Student Center.
  • Make the route cycling instead of walking, from the Memorial Student Center to the Century Tree.
  • Give me walking directions from the Memorial Student Center to the Century Tree.
  • Find a nearby park I can walk to in College Station.
  • Give me a route I can finish in under an hour, starting at the Memorial Student Center.
  • Give me a route to the closest coffee shop from Texas A&M, College Station.
  • What’s the best walking route to campus landmarks from Texas A&M, College Station.

Assumptions

  • Calorie burn is approximated with a fixed value in code (e.g. 50 kcal/km for walking). This is to keep the MCP logic simple and predictable.
  • When the user does not provide origin/destination, the server falls back to the defaults from .env.
  • Google Maps API must be enabled in the Google Cloud project for the calls to succeed.

Troubleshooting

  • If the response contains "error": "REQUEST_DENIED" or similar, check:

    • API key is present
    • Billing is enabled
    • Directions/Geocoding/Places are enabled

  • If the response is empty, ensure .env is loaded and the server was started in the same environment.

Share this project:

Updates