Quickstart
Five minutes from zero to a live app with a database, auth, and a deploy URL you can share.
Install Atelier Desktop, sign in, and connect at least one LLM provider (Claude, ChatGPT/Codex, or local). Quickstart assumes Studio is open.
1. Create a project
In Studio, click New project and pick the React + Vite starter. Atelier scaffolds your repo with a working preview, @atelier/sdk preconfigured, and an empty migration directory.
2. Tell it what to build
In the chat panel, describe what you want:
Build a todo app. Items have a title and a done flag. Sync across devices.
Studio plans the change, writes a migration, generates the UI, and shows a preview. Realtime sync is wired up because Base ships it by default — you don’t have to ask.
3. Inspect what Base did
Open the Base console (sidebar → Base) to see what was created:
- A
todostable with RLS policies tied to the signed-in user - A storage bucket (unused for now, but ready)
- End-user auth wired to Google by default
- A realtime channel that broadcasts row changes to subscribers
4. Read and write from your code
The generated code uses @atelier/sdk:
import { atelier } from '@atelier/sdk';
const todos = await atelier.from('todos').select('id, title, done');
await atelier.from('todos').insert({ title: 'Ship the thing' });
5. Deploy
Click Deploy. Atelier builds the bundle server-side, applies pending migrations to production, uploads, and rolls out a canary:
https://<your-app>.apps.atelier.style
The rollout starts at 5%, watches error rate and P95 latency, then steps to 25 / 50 / 100%. Any regression rolls back automatically.
6. Share it
Send the URL to anyone. The first visit triggers Base auth (Google by default). Their session cookie is scoped to .apps.atelier.style, so signing into one of your apps signs them into all of them.
Next
You now have the core loop. From here, dive into specific primitives:
- Database — schemas, RLS, queries, migrations
- Auth — providers, sessions, cross-app SSO
- Realtime — channels, broadcast, presence
- Functions — server-side logic
- Vector Search — embeddings, similarity, RAG