In Prototo Desktop, open your project and you are already there: your coding agent (Claude Code or Codex) is running in the left pane, your iPhone is live on the right.
Picked Codex? The first time it opens a project it asks whether you trust the folder. Say yes so it can use Prototo’s preview tools.
In the terminal flow, open a second Terminal window and navigate to your project:
cd myapp
claude # or: codexEither way, your agent is now running inside your Proto project. It has already read your DESIGN.md and AGENTS.md. It knows your design system, your component library, and the rules for generating screens. You don’t repeat yourself.
Describe your first screen
Type this into the agent:
add a settings screen with notification and dark mode togglesWatch the Simulator. In about 10 seconds your screen appears.
Real native layout. Real scroll physics. Your design system applied automatically — colour, typography, spacing, shape, all pulled from DESIGN.md.
Iterate with prompts
Keep describing. Your agent keeps generating.
make the quick action buttons pill-shaped with an icon
above the label, arranged in a row with equal spacingadd a bottom nav with Home, Cards, Activity and Profile tabsadd a profile screen with a large avatar, the user's
name and bio, and a grid of recent postsEach prompt updates the Simulator in seconds. You never touch a file.
What just happened
The flow is the same every time:
1. You type a prompt in your agent
2. Your agent reads DESIGN.md and AGENTS.md for context
3. Your agent writes a new screen file to /screens/
4. The Simulator hot-reloads — the screen appearsTotal time: ~10 seconds.