AndesCoach turns raw Garmin data into actionable, science-backed coaching advice. Built around modern sports science — TSS, CTL/ATL/TSB, ACWR, HRV, NGP — and orchestrated by a large language model that knows your training history.
Six tightly integrated modules, one daily briefing.
CTL, ATL, TSB and ACWR computed nightly with EWMA, segmented by sport, with injury-risk zones surfaced in real time.
Per-activity Normalized Graded Pace (Skiba 2012 + Minetti 2002) and tiered TSS — Power, Running, Heart-rate or sRPE depending on data quality.
HRV, RHR, sleep and Body Battery streamed from Garmin, baselined against your own 7-day rolling history.
Daily caloric balance with MFP-style food diary, weight tracking, and macros — calibrated against measured expenditure.
Every morning, a 60-word briefing in your language — based on planned workout, executed history, readiness and yesterday's load.
Versioned physiology snapshots (LTHR, FTP, HRmax) so every past activity is scored against the thresholds you actually had that day.
AndesCoach started as a personal project: an endurance athlete based in Santiago, Chile, wanted a single tool that combined the rigor of TrainingPeaks with the responsiveness of Garmin Connect and the contextual intelligence of a modern large language model.
Existing platforms were either too rigid (fixed thresholds, no AI), too generic (no trail-specific calibration, no NGP), or too expensive for a single athlete to justify. So we built our own.
Today AndesCoach runs in private beta. The roadmap includes opening it to a small group of trail and ultra athletes during 2026.
A Spring Boot 3 backend (Java 21) ingests Garmin data and computes sports-science metrics. A React + Vite frontend visualises everything in an Apple-inspired interface. MongoDB stores activities, profiles, and AI briefings. The Anthropic API powers the coaching agent.
AndesCoach is currently in private beta. If you're a trail or endurance athlete interested in joining the early access list, drop us a line: