RaptorCode Studio

1. Describe the app on the home page or in a new project. Optionally attach a text brief (.txt, .md, .csv, .json, .html, up to 64 KB).

2. RaptorCode applies your first prompt to a pinned Next.js + TypeScript + Material UI + Motion starter and streams the changed files. You see the output arrive live in the Code tab.

3. The files are validated (paths, imports, fixed dependencies, size) and merged into a complete snapshot, saved as a new version and shown in the preview.

4. Keep chatting to change it. Each request sends the current project source plus recent conversation as context; untouched files are kept.

5. Open History to restore any earlier version. Restoring creates a new version, so nothing is lost.

6. Export downloads a .zip of the full Next.js project. Run npm install, npm run dev or npm run build (Node.js 20.9+). Older HTML projects still export index.html.

Output is a Next.js App Router project in TypeScript using Material UI (Emotion), plain CSS and Motion (motion/react). package.json, tsconfig, next.config, the root layout and providers are managed and cannot be changed by prompts; dependencies are fixed.

The in-builder preview compiles and runs the client UI in an isolated sandbox with network blocked. It is not a Next.js server: server rendering, server routes and metadata only run after export.

Backends, databases, server routes, real authentication and payments inside generated apps are not supported; apps use sample data. MUI icons and extra npm packages are not available.

RaptorCode does not deploy or host generated apps.

Generated code is treated as untrusted. The preview iframe uses sandbox="allow-scripts" without allow-same-origin, so it gets an opaque origin and cannot read your session, cookies or storage.

A Content-Security-Policy is injected that blocks fetch/XHR/WebSocket, form submission, framing and remote scripts.

Projects, messages and versions are protected by owner-only access rules. The generation endpoint verifies your session and project ownership server-side; AI keys never reach the browser.

Prompt: 4000 characters. Attached brief: 20,000 characters.

Output is stopped above 250,000 characters. One generation at a time per account.

AI app builds, brief improvements and next-edit suggestions share the same rolling hourly and 24-hour request allowance. Failed and cancelled attempts also use a request. Cached suggestions use no additional request. Completed app counts exclude helpers.

Paid subscriptions raise your AI request limits once payment is confirmed. Cancel anytime from Billing.

If a generation is cancelled or the model returns an incomplete document, no version is saved.