Be my trainer and nutritionist across my watch, food log and scale
An Italian user connected the AI over MCP to three sources of truth — Garmin for activity and recovery, Yazio for food, a FitDays scale for weight and body composition — and gave it a coach's rulebook in CLAUDE.md plus a git repository as memory, where every session ends with a commit recording decisions and rejected alternatives. Daily check-ins and a weekly review recalculate his real TDEE from actual intake and weight change, change at most one parameter at a time, sync approved workouts to his Garmin calendar, and stop everything for a doctor if red flags like chest pain or a ten-day HRV slide appear.
How it works
- Connect Garmin, Yazio and FitDays through MCP servers, one domain each
- Write the coach's rules in CLAUDE.md: one active goal with an exit criterion, calorie floor, protein range, mandatory deloads
- Keep memory in a git repo: session diary, learnings, decisions with rejected alternatives, append-only data CSVs
- Run daily check-ins (weight, food, sleep, HRV in three numbers and one action) and weekly reviews that recompute TDEE and change at most one parameter
- Confirm every write before it syncs a workout to Garmin or logs food
Why it’s worth knowing
Its memory is a git repo: every coaching decision — and every rejected alternative — is committed, so the coach can never quietly re-propose what he ruled out.
Where it’s from
Summarized from a public post on dev.to. See the original ↗