TrainingPeaks MCP Server

MCP server for TrainingPeaks. Reads workouts, fitness metrics (CTL/ATL/TSB), and health data. Creates, updates, and deletes planned workouts. Built with Python/FastMCP with browser-based auth and auto-refresh. Designed for AI-powered training plan management and automated adjustment based on performance, fatigue, and injury.

Setup

python3 -m venv .venv
source .venv/bin/activate
pip install -e .

Related MCP server: claude-garmin

Usage

python -m training_peaks_mcp

On first run, a browser window opens with the TrainingPeaks login page. Enter your credentials — they're sent to localhost only, saved locally to ~/.config/training-peaks-mcp/credentials.json, and auto-refresh when expired.

CLI Options

python -m training_peaks_mcp              # Start the MCP server
python -m training_peaks_mcp --auth       # Re-authenticate
python -m training_peaks_mcp --auth-status # Check auth status
python -m training_peaks_mcp --auth-clear  # Remove stored credentials

Tools

Auth

Read

Write

Authentication

Three methods, checked in order:

1. Stored credentials — saved automatically after first login, auto-refreshes on expiry

2. Browser login — opens on first run if no credentials exist

3. Environment variable — set TP_AUTH_COOKIE to override (for CI/containers)

Credentials are stored at ~/.config/training-peaks-mcp/credentials.json with 0600 permissions.

Architecture

src/training_peaks_mcp/
├── __init__.py
├── __main__.py      # CLI entry point
├── auth.py          # Login flow, credential storage, browser auth page
├── client.py        # Async HTTP client for tpapi.trainingpeaks.com
└── server.py        # FastMCP tool definitions

The server uses the TrainingPeaks internal API (tpapi.trainingpeaks.com). Auth cookies are exchanged for short-lived bearer tokens via /users/v3/token.

Supported Sports

Swim, Bike, Run, Brick, Crosstrain, Race, DayOff, MtnBike, Strength, Custom, XCSki, Rowing, Walk, Other

License

MIT