Learn fullstack development. Prototype with AI. Keep total control.
Most frameworks are too complex for AI. They hide logic behind "magic" and deep abstractions. This causes AI agents to hallucinate and break things.
We built StartER to stand out. It is a "Zero-Magic" foundation designed for Human-AI co-creation. By keeping the code readable and explicit, we provide AI agents with a perfect mental model. This makes it the ultimate playground for rapid prototyping and learning.
This project shows a simple and readable way to structure a fullstack app with:
- Express backend
- React frontend
- shared contracts for API
If you are looking for a "Express + React starter" or "Node React boilerplate", this repository is a practical example.
# 1. Clone the playground
git clone https://github.com/rocambille/start-express-react.git my-prototype
cd my-prototype
# 2. Install dependencies and initialize the database
npm install
cp .env.sample .env
npm run database:sync
# 3. Start co-creating
npm run devDon't let AI "guess" your architecture. Use make:clone to replicate working logic.
npm run make:clone -- src/express/modules/item src/express/modules/task item taskThis enforces consistency by cloning your actual code patterns. This keeps your AI agent focused and accurate.
You define API behavior in tests/contracts.ts: a central, declarative source of truth.
- For you: clear, living documentation.
- For AI: a strict "contract" it must follow when generating endpoints.
- For the app: instant verification that the AI didn't miss a scenario.
- Sync SQLite: direct data access that AI can read and write without
async/awaitconfusion. - Explicit casting: we verify data at the edge. This prevents the silent bugs AI often introduces.
- Transparent stack: Express 5 + React 19. No black boxes. You understand every line.
- Backend: Node.js, Express 5, Zod (validation)
- Frontend: React 19, React Router, Vite, Pico CSS
- Database: SQLite (zero-config, sync API)
- Tooling: TypeScript, Biome, Vitest, Docker
StartER is a pedagogical project. If this architecture helps you prototype faster with AI, give us a ⭐ on GitHub!
👉 Read the wiki & AI mental model
Distributed under the MIT license. You are free to use, modify, and redistribute it for educational or professional purposes.
