IRON.

A fitness tracker built for the set between sets. React Native · FastAPI · PostgreSQL · MCP.

---

Why

Logging a set into a commercial fitness app takes 20–40 seconds of tapping. Between sets you have 60–90. IRON is designed so that logging takes 10 seconds, max, every time.

Four design principles hold the line:

1. Data-first — every pixel serves the measurement. No decorative UI.
2. Noir + Acid — deep black background, acid green (#BFFF00) only where it matters (the last number you care about).
3. Mono for data — JetBrains Mono for every measurable value; Bebas Neue for titles; Outfit for body.
4. Log in 10 seconds — tap-tap-tap, no modal nesting, no confirmation popups between sets.

Stack

Component	Choice	Why
Mobile	React Native + Expo (Expo Router)	File-based nav, fast iteration
Backend	FastAPI + SQLAlchemy + PostgreSQL	Typed, fast, boring
Auth	Apple Sign-In + Google OAuth + JWT	Mobile-native, no email/password
Health data	Apple HealthKit bridge	Bodyweight syncs automatically
AI	OpenAI (exercise suggestions, muscle mapping)	Targeted, not chat
MCP server	Python · FastMCP	Lets Claude read/write your training data
Hosting	Railway (API + Postgres)	git push to prod

Features

• 🏋️ Sessions & sets — log a workout set by set (weight · reps · RPE), with templates for repeat sessions
• 📈 Progress — per-exercise history, 1RM (Epley/Brzycki), trend over time
• 🫁 Body sync — bodyweight pulled from Apple HealthKit, no manual entry
• 🧠 Smart exercise suggestions — AI proposes next logical exercises from your split and recent volume
• 🎯 Muscle taxonomy — every exercise is tagged with primary and secondary muscles, drives visualizations
• 🔌 MCP integration — point Claude at your IRON API and ask "how's my bench trending?" from your editor
• 👥 Multi-user — full data isolation via user-scoped foreign keys

Architecture

┌────────────────────────┐
│  React Native (Expo)   │
│  • (tabs) Lift/History │
│  • Session detail      │
│  • Healthkit bridge    │
└───────────┬────────────┘
            │ HTTPS + JWT
            ▼
┌────────────────────────┐         ┌──────────────────────┐
│  FastAPI API           │         │  MCP server          │
│  • /auth (Apple/Google)│         │  • read/write tools  │
│  • /sessions · /sets   │◀───────▶│  • same DB           │
│  • /bodyweight         │         │  • for Claude /      │
│  • /ai (suggestions)   │         │    AI assistants     │
└───────────┬────────────┘         └──────────────────────┘
            │
            ▼
┌────────────────────────┐
│  PostgreSQL            │
│  users · exercises     │
│  sessions · sets       │
│  bodyweight · templates│
└────────────────────────┘

Project structure

.
├── app/                 React Native (Expo Router)
│   ├── app/             file-based routes (tabs, session/[id], exercise/[id])
│   ├── components/      StatCard, SessionCard, ExerciseLog, SetRow, ProgressChart, ...
│   ├── lib/             api client, theme, healthkit bridge, 1RM math
│   └── assets/          fonts, icons
├── api/                 FastAPI backend
│   ├── main.py          app entry, lightweight live migrations
│   ├── models.py        SQLAlchemy (User, Exercise, Session, Set, Bodyweight, ...)
│   ├── schemas.py       Pydantic
│   ├── auth.py          JWT + Apple + Google verifiers
│   └── routes/          auth, sessions, bodyweight, ai
├── mcp/                 MCP server (Python · FastMCP)
│   └── server.py        tools that talk to the same API
├── data/                sample session data (weights in kg)
├── docs/plans/          design & implementation docs
├── Muscles-iron.svg     anatomy sprite used by the muscle visualizer
└── railway.toml         deploy config

Getting started

Backend

cd api
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env     # set DATABASE_URL, JWT_SECRET_KEY, GOOGLE_CLIENT_IDS
uvicorn main:app --reload

Mobile app

cd app
npm install
npx expo start

MCP server (optional)

cd mcp
pip install -r requirements.txt
export IRON_API_URL=http://localhost:8000
python server.py

Then wire it into Claude Desktop or your MCP-capable client.

Required environment variables (API)

Variable	Purpose
DATABASE_URL	Postgres connection
JWT_SECRET_KEY	32+ chars, openssl rand -base64 32
GOOGLE_CLIENT_IDS	Comma-separated Google OAuth client IDs
OPENAI_API_KEY	Smart exercise suggestions (optional)

Apple Sign-In is verified against Apple's public JWKS; no client secret needed.

Status

Single-developer project, extracted from personal use. The backend has been multi-user since day ~30 — data is scoped by user_id across all resources. The MCP layer is the more unusual piece: it lets an AI assistant treat your training data as a first-class resource.

License

MIT

---

Built by @thomasetienne