AI-powered fitness coaching with natural language
Kinetic is an AI personal training app that provides adaptive, conversational coaching for lifters. Users can chat naturally with their AI coach instead of filling out forms.
- Frontend: Next.js 14 + TypeScript + Tailwind CSS
- Backend: Express + Prisma + MongoDB
- AI: OpenAI API (GPT-4o)
- Auth: Simple user system (no email required for demo)
- Natural language AI coaching
- Adaptive workouts based on daily check-ins
- Custom exercise library with media (images, videos)
- User dashboards with progress tracking
- Workout logging and PR tracking
kinetic/
├── SPEC.md # Full project specification
├── docs/
│ ├── onboarding-flow.md # 14-step user onboarding
## Icons
This project uses [better-icons](https://github.com/better-auth/better-icons) for AI-powered icon search.
### Using with Codex
The project has an MCP config at `mcp.json` — Codex can search and add icons using natural language:
- `"Search for fitness icons"`
- `"Get lucide:dumbbell as SVG"`
- `"Add a workout icon to my dashboard"`
│ ├── exercise-library.md # 75+ exercises database
│ └── ai-prompts.md # AI coaching prompts
├── src/ # Next.js frontend
│ ├── app/
│ │ ├── page.tsx # Landing page
│ │ ├── demo/ # Demo with sample workout
│ │ ├── dashboard/ # User dashboard
│ │ ├── workouts/ # Workout creation
│ │ └── exercises/ # Exercise creation
│ └── lib/
│ └── api.ts # API client
└── backend/
└── src/
├── routes/ # API endpoints
├── services/ # Business logic
└── prisma/ # Database schema
# Frontend
cd kinetic && npm run dev # http://localhost:3000
# Backend
cd kinetic/backend && npm run dev # http://localhost:3001| Endpoint | Description |
|---|---|
GET /api/health |
Health check |
GET/POST /api/users |
User management |
GET /api/users/:id/dashboard |
Dashboard data |
GET/POST /api/exercises |
Exercise library |
GET/POST /api/workouts |
Workout CRUD |
POST /api/ai/generate-workout |
AI workout generation |
POST /api/ai/chat |
Chat with AI coach |
See docs/ for detailed research:
- Competitor Analysis: Trainerize, FitnessAI, Zing Coach, Humango, Strong, JEFIT
- AI Prompt Engineering: Two-stage generation, persona system, safety guardrails
- Dashboard UX: Best practices from top fitness apps
- Clone the repo
- Create a branch
- Make changes
- Submit PR
MIT