Training that adapts to your data, not a calendar

Five progression methods, all first-class

Double progression (target rep range, fixed sets, load increases when the top of the range is hit), linear (fixed load increment per session), RPE (autoregulated load from rate of perceived exertion), percentage (% of estimated 1RM, recalculated per Epley), and wave loading (cycle through intensity blocks).

Pick one or let your coach configure it per exercise. Switch any lift to a different method without rebuilding the program. The engine is deterministic — every progression decision is explained: "increased squat to 105 kg because you hit 5 reps at 100 kg twice in a row at RPE ≤ 8."

Reactive deload, not calendar deload

Most apps deload every 4-6 weeks regardless of how training is actually going. That's a guess. MuscleBuddy tracks per-exercise stall signals (missed reps, consistent RPE drift, recovery score trends) and triggers a targeted deload only on the lifts that need it.

If just one lift is stalling, only that lift deloads. If everything's stalling — likely a recovery issue — a full block deload kicks in. Pre-planned deload remains available if you prefer that style, but reactive is the default.

Equipment-aware program generation

Tell MuscleBuddy what equipment you have access to (or your gym does). Every program generated filters to lifts you can actually perform. No "barbell back squat" prescriptions when your home gym is dumbbells-only.

The exercise KB ships with 45+ entries (and growing) — each tagged with primary muscle, secondary muscles, equipment requirements, and a difficulty bucket. Substitutions are suggested automatically when an exercise isn't supported.

Built on a pure rule engine

The training math lives in `@musclebuddy/engine` — a TypeScript package with zero database or HTTP dependencies. Every progression decision can be unit-tested, reproduced offline, and replayed against historical data. There's no "the AI decided to deload" black box.

Coaches can override anything via per-client `CoachConfig`. The engine produces a suggestion; the coach approves, modifies, or rejects.

Frequently asked

Can I use my own program?

Yes. The progression engine works on any program — log sets, the engine handles load adjustments. You can also generate a fresh program from one of 5 templates.

Is the AI generating my workouts?

No. Workout generation is rule-based (split-calculator + exercise-kb). LLMs are only used for conversational features like check-in summarization, never for the core training logic.

What methods does MuscleBuddy support?

Double progression, linear, RPE, percentage, wave, and manual override. Configurable per exercise.

How does deload work?

Reactive by default — only triggered when training data shows stalling. Pre-planned deload (e.g., every 5th week) is available if you prefer that style.

Does the training engine know my gym's equipment?

Yes. Set your gym profile once; programs filter to supported equipment automatically.