MuscleBuddy
An AI-adaptive training and nutrition platform with deep analytics, a full health module, and a social gamification layer
Features
- •AI-adaptive training — five progression methods (double, linear, RPE, percentage, wave loading), automatic deloads, and equipment-aware generation from a 300+ exercise knowledge base
- •Microcycle scheduling that plans around your real week, mesocycles, and deload timing instead of a fixed calendar grid
- •Adaptive nutrition — Mifflin-St Jeor TDEE with calorie and macro targets that auto-adjust off a rolling weigh-in trend, plus carb cycling, reverse dieting, and contest-prep modes
- •Nutrition tooling — a 7,000+ item food knowledge base backed by USDA and Open Food Facts, custom foods, recipes, and a meal planner
- •Pro analytics — correlational analytics, bloodwork correlation, LLM-narrative reports, per-lift Lift Reports, trend tracking, and a training-volume heatmap
- •Health module — recovery scoring, bloodwork with reference-range flags, vitals tracking, and a supplement and compound interaction checker
- •Ranks and gamification — a nine-tier muscle-rank system, nutrition ranks, XP levels, streaks, achievements, and leaderboards
- •Social layer — an activity feed, follows, and user discovery, plus an AI coach chat for conversational guidance
- •Offline-first PWA — workouts and food logs queue in IndexedDB and flush on reconnect via a Workbox service worker
- •Stripe billing with Stripe Connect coach revenue share for coaches and gyms managing clients
Screenshots
What it is
MuscleBuddy is a fitness coaching PWA built on the premise that most training apps are either too simple to be useful or too complicated to stick with. It serves the full range from first-time gym-goers to competitive bodybuilders from a single codebase, using configuration and progressive disclosure to surface only what each user actually needs.
The core of the product is an algorithm-first engine. Progression calculations, TDEE estimation, macro targets, and nutrition auto-adjustment are deterministic rule engines — not AI. Anthropic's Claude handles conversational features, check-in analysis, and coaching explanations, but no core calculation depends on the LLM. That separation keeps the platform fast, consistent, and fully testable.
The stack is a TypeScript monorepo: a React 19 + Vite frontend, a Fastify backend, PostgreSQL via Prisma, and Supabase for auth and storage. State is managed with Zustand, the UI is styled with Tailwind, and offline support runs on Dexie/IndexedDB with a Workbox service worker. Stripe and Stripe Connect handle billing and coach revenue share, Nutritionix and Google Places power food and gym lookups, and per-post Open Graph cards are generated server-side with satori. The progression, nutrition, ranking, and health logic all live in a shared packages/engine directory as pure, deterministic TypeScript with no I/O dependencies.
How it works
A user sets their goal, schedule, and available equipment during onboarding, and MuscleBuddy generates a full training program — exercise selection, set and rep schemes, a progression method, and a deload schedule — bounded so every session is coherent rather than a one-exercise stub. The schedule is a microcycle planned around the user's actual week rather than a fixed calendar grid. Each session, the app tracks completed sets and logged weights to decide whether to advance load, hold, or trigger a deload. Weigh-ins — logged by hand or synced from a Withings smart scale — feed a rolling average that drives automatic calorie adjustments toward the user's goal without manual recalculation.
Around that core sits a fuller platform. Adaptive nutrition supports carb cycling, reverse dieting, and contest prep on top of a 7,000+ item food knowledge base. A Pro analytics suite adds correlational analytics, bloodwork correlation, LLM-narrative reports, per-lift Lift Reports, trend tracking, and a volume heatmap. A health module covers recovery scoring, bloodwork with reference-range flags, vitals, and a supplement and compound interaction checker. A ranks-and-gamification layer — nine-tier muscle ranks, nutrition ranks, XP levels, streaks, achievements, and leaderboards — sits alongside a social layer of feeds, follows, discovery, and an AI coach chat.
The same PWA installs on iOS and Android without an app store, and ships with published privacy and terms, account data export and deletion, and medical disclaimers on the health-related screens.