I built Pace, an MCP server that pulls your wearable data into Claude so you can actually talk to it.
You connect your devices once (Garmin, Oura, Whoop, Polar, Fitbit, 20+) and Claude gets 15 tools: sleep analysis, workout detail with lap splits, HR zone breakdown, recovery status, 30/90-day trends, and more.
The thing that surprised me most: Claude's artifact/visualization feature turns this into something way more useful than I expected. You ask "analyze my last 10 runs" and you get an interactive chart of pace trends, HR zones per run, and a written summary — all in one response. No dashboard needed.
Tools available right now:
- get_sleep_data — deep sleep, REM, HRV, SpO2, sleep score
- get_workout_detail / get_workout_laps / get_workout_samples — full session breakdown
- get_recovery_status — readiness, resting HR trend, HRV baseline
- get_training_summary — load, volume, consistency over time
- get_trends — any metric over 7d/30d/90d with baseline comparison
Stack: Python (FastMCP) · Cloud Run · Cloud SQL PostgreSQL · Firebase · Terra API · OAuth 2.1
It's live and free to try at pacetraining.co. Claude Pro/Max not required (custom connector).
Would love feedback, especially from people who've built MCP servers or use wearables seriously.
