Build a glasses app from a prompt: how GlassKit Studio works
GlassKit Studio turns a text prompt into a working, publishable Meta Ray-Ban Display app: plan, generate, preview in a virtual glasses display, edit by chat, and publish to a URL with a QR code.
TL;DR: GlassKit Studio is a prompt-to-app builder for the Meta Ray-Ban Display. You describe the app; Studio plans it with you, generates a working React app on the free GlassKit UI library, shows it running in a virtual 600 × 600 glasses display, lets you refine it by chat, and publishes it to a stable HTTPS URL with a QR code. Prepaid credits, no subscription.
The flow, end to end
- Plan. A new app starts in plan mode: Studio converses about what you're building and proposes a concrete plan before spending a build. You can attach reference images ("make it look like this"), sprites and sounds to use in the app, or documents for context, and dictate by voice.
- Build. One click turns the plan into a complete glasses app:
real React + TypeScript on
@glasskit-ui/react, already following the platform rules (600 × 600, additive dark design, D-pad focus,.focusablenavigation). Attached sprites and audio are mounted as app assets the code can use. - Preview. The app runs immediately in a virtual glasses display in your browser: arrow keys stand in for the Neural Band, exactly as they do on-device. If the generated code fails to build, Studio detects the error and repairs it automatically.
- Refine. Edits are conversational ("make the timer bigger", "add a pause screen"). Studio applies focused changes to the existing code rather than regenerating the app.
- Publish. One click hosts the app at a stable HTTPS URL
(
yourapp.glasskit.studio) and generates a QR code that encodes it. Scan it with the Meta AI app (developer mode on) and the app is on your glasses. That matches Meta's official distribution model, where Web Apps load from a public HTTPS URL (wearables.developer.meta.com); see How to publish.
What generated apps can use
Everything the platform gives Web Apps, through GlassKit UI's typed hooks:
| Capability | In a Studio app |
|---|---|
| D-pad + select input | useDpad() focus over .focusable elements |
| Screens and back | useNavigator() stack navigation |
| Motion / orientation | useDeviceMotion() / useDeviceOrientation() |
| Location | useGeolocation() (via the paired phone) |
| Sprites, images, sounds | Your attached assets, mounted and importable |
| Persistence | localStorage (on-device) |
Studio apps are client-only by design: no camera or microphone (a Web Apps platform limit), and no server. The moment your idea needs accounts, payments, or shared data, that's the eject to Stack moment.
Pricing
Studio is pay-as-you-go: prepaid, non-expiring credits, no subscription. 100 credits = $1. New accounts get free welcome credits to build with; top-up packs run $10 / $25 / $50 / $100, and every GlassKit Stack purchase includes 2,500 Studio credits.
When to eject
Studio's ceiling is deliberate: it builds superb on-device apps, not backends. Eject to Stack takes your generated app and drops it into the production monorepo, where it gains a Next.js companion site, Clerk auth, payments, and a Convex backend without a rewrite. See GlassKit Stack vs building from scratch.
Try it now
The prompt box at glasskit.app/studio is the product. Describe an app; you'll be watching it run in the virtual display a minute later.