Personal ToolAI-ManagedMobile PWASmithClaude IntegrationSolo BuildFitness

Gym Routine Builder

A personal gym companion app I built just for myself. Mobile-first PWA that manages 8 custom routines across 33 machines at my gym, tracks every set and rep in real time, and logs workout history with progress metrics. The twist: routines are created and managed by SmithClaude via conversational commands, making it the first app in my portfolio where an AI agent is both the builder and the content manager.

View live

The Problem

Generic gym apps are bloated with features you never use, filled with exercises you don't have equipment for, and impossible to customise to your specific gym's machines. I train at a gym with 33 specific machines and needed an app that knew exactly which ones were available, could build routines tailored to my goals (hypertrophy, golf mobility, specific muscle groups), and tracked progress without the noise. I also wanted to manage routines conversationally — tell SmithClaude 'build me a chest day' and have it appear in the app, ready to use.

Key Metrics

8 routines, 0 manual setup

AI-created content

33 machines catalogued

Gym coverage

Built in a weekend

Speed to production

Conversation to workout

SmithClaude integration

My Role & Approach

I built a focused, mobile-first PWA designed for exactly one user and one gym. Instead of building a routine editor UI, I connected the Supabase backend directly to SmithClaude — routines are created, edited, and optimised through natural conversation via Telegram or terminal. SmithClaude knows every machine at my gym, understands muscle groups and training principles, and writes routines with proper exercise order, set/rep schemes, and notes. The app itself is pure execution: pick a routine, start the workout, log your sets, track your progress. No fluff, no social features, no discovery — just train.

Tools & Tech

Development

Next.jsReact

Other

TypeScriptTailwind CSSSmithClaude (routine management via Supabase REST API)

Databases

Supabase

Deployment

Vercel

The Build Process

Research & Setup

I catalogued all 33 machines at my gym with categories (upper body, lower body, core), brands, and descriptions. Set up Supabase database with tables for machines, routines, routine items (with exercise order, sets, reps, notes), workout sessions, and workout sets.

SmithClaude Integration

I connected SmithClaude to the gym database via Supabase REST API. The workflow: I say 'create a back day routine', SmithClaude queries available machines, designs an optimal routine based on muscle groups and training principles, and inserts it directly into the database. Routines appear in the app instantly.

App Build

I built a dark-themed, mobile-first PWA with four core screens: Today (dashboard with routines and recent workouts), Routines (browse and start), Machines (equipment guide with search), and Metrics (progress tracking with volume, duration, streaks, and per-machine progression).

Workout Tracking

I built the real-time workout logger: start a routine, see each exercise with target sets/reps, log actual weight and reps per set, track rest times, and save the complete session. All data flows into metrics for long-term progress tracking.

Learnings

“Building for one user is liberating — every decision is faster because there's no debate about user personas or edge cases. The SmithClaude integration proved that AI agents are most powerful when they manage the data layer of apps, not just the UI. I never open an admin panel to create routines — I just tell SmithClaude what I want and it appears in the app. This pattern (AI as content manager, app as execution layer) is something I want to explore in client projects. The dark theme and mobile-first approach were non-negotiable for a gym environment — bright screens and tiny buttons don't work when you're between sets.”