STRUGGL: Zero-LLM-Math AI coach for hybrid athletes — Strength, Running & Hyrox

Hey everyone — Taha here, founder of STRUGGL.

Why I built it I train strength + running + Hyrox while shipping code all week. Every time I asked an AI assistant to program a hybrid week, it quietly botched the math: 220-age heart-rate zones off by 12+ bpm, weighted pull-up percentages calculated on belt weight alone instead of bodyweight + belt, or compromised Hyrox run splits treated like fresh track 10Ks.

What STRUGGL does

• Enforces a Zero-LLM-Math rule: Claude handles natural-language coaching, schedule constraints, and session adjustments, while 8 deterministic Python engines compute every load, Zone 2 bpm, 17-muscle MEV/MAV/MRV ledger, 7:28 EWMA ACWR fatigue ratio, and Hyrox station/lap split.
• Works across Web, iOS & Android, and exposes a remote OAuth 2.1 MCP server (struggl.singandclap.com/mcp) + open-source physiology primitives (github.com/Kubyer/struggl-mcp).

What I'd love feedback on

1. For fellow hybrid athletes / lifters / runners here: how does the Claude MCP coaching loop feel on your real weekly block?
2. For builders: any edge cases you'd want added to our open-source physiology primitives or fitness-math-auditor subagent?