Skip to content

About

Open-source AI interface builder: drag components onto a canvas, ask an AI to build pages, export HTML.

Topics

Resources

Stars

3 stars

Watchers

1 watching

Forks

Latest commit

 

History

103 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

openconstructor — build interfaces with words and blocks

openconstructor

Build interfaces with words and blocks — an open-source, in-browser AI interface builder.

openconstructor.pw — try it live

Roblox UI builder · AI UI generator · 320+ components · HTML to React / Vue / Svelte · LÖVE 2D UI · Gallery


Drop ready-made components onto one shared canvas, move them around freely, and ask an AI model to build or change anything on the page. When you're done, export a single self-contained HTML file.

Features

  • Component catalog — 320+ components in 13 sections (AI & Chat, Dashboard, Commerce, Auth, Landing…), all plain HTML + CSS. Animated ones play on hover.
  • Free-form canvas — drag, pan, zoom, desktop / laptop / tablet / phone sizes.
  • Background editor — solid, linear, radial and mesh gradients, dot / grid / line / noise patterns, or raw CSS.
  • ⚗ Fusion — right-click an element → fuse with… → click another: the AI breeds a new component that does the first one's job in the second one's visual style.
  • AI assistant — 20 free generations a day when signed in (GPT-OSS 120B via a small server proxy), or bring your own key for OpenRouter, Google Gemini, Groq, Cerebras, Mistral, DeepSeek, OpenAI, Anthropic, xAI or Together — those keys stay in your browser and go only to the provider you choose. Paste a screenshot to rebuild it with a vision model.
  • Real photos & attachments — the AI fills sites with stock photos (CC0, found by what should be in the picture) instead of grey boxes; press + or paste/drop to attach your logo, product photos or a text file to a request.
  • 3D models — ask for a 3D object or scene: the AI describes it as data (shapes, materials, lights, motion) and a built-in Three.js renderer draws it — animated on the canvas, orbit it in ▶ play, download it as .glb for Blender, Unity, Godot or Roblox Studio.
  • Edit without code — double-click to edit text in place, restyle any part, swap in your own image, tune every animation (speed, direction, repeats).
  • Multi-select — Shift-click or Shift-drag a box, move together, align & distribute; smart guides while dragging.
  • Screens & prototypes — menu → shop → inventory: several screens in one project, link any button to #screen and it opens that screen in ▶ play, in the published page and in the Roblox export.
  • Roblox Studio sync (guide) — install the plugin, press Export → Sync to Roblox Studio, paste the code in Studio: the UI lands in StarterGui with named buttons. Live mode re-syncs on every change, and your own button scripts are kept.
  • Game UI for Roblox — 25 HUD components and 5 full screens (simulator, tycoon shop, shooter, obby, mobile RPG), Roblox safe-zone overlay, any page size from all four edges.
  • Publish & share — one-click public link, community gallery with remix, embed code, cloud projects, version history.
  • VS Code–style layout — resizable, movable, collapsible panels.
  • Safe previews — every element renders in a sandboxed iframe with scripts and network requests disabled.
  • Export anywhere — HTML, React, Vue, Svelte, plus Roblox (Luau LocalScript in pixel or scale mode, or a drag-and-drop .rbxmx model with named buttons and click handlers) and LÖVE 2D (Lua): the rendered layout is converted into native UI code. AI-assisted ports to Flutter, SwiftUI, Jetpack Compose and Tailwind.
  • Optional sign-in with Google, GitHub or email (Supabase).

Quick start

git clone https://github.com/Kuloka/openconstructor.git
cd openconstructor
npm install
npm run dev

Open http://127.0.0.1:5173. Press ? in the editor for a quick tour.

To use the AI assistant, click the gear icon and paste a key from any supported provider. Google Gemini and Groq are the quickest to set up.

Configuration (optional)

Copy .env.example to .env.local:

Variable Purpose
VITE_SUPABASE_URL, VITE_SUPABASE_ANON_KEY enable accounts (Google, GitHub, email)
VITE_DONATE_URL link for the "support" button
VITE_CRYPTO_WALLETS `coin

Without these, the app works fully, just without accounts.

Build & deploy

npm run build   # output in dist/

npm run deploy builds and uploads to Cloudflare (config in wrangler.jsonc). dist/ is a plain static site, so Netlify, Vercel or GitHub Pages work too.

Stack

Vite · vanilla JS editor · React 19 for the welcome screen and chat input · Supabase Auth

Contributing

Issues and pull requests are welcome. New catalog components go in presets*.js: self-contained HTML + CSS, a unique class prefix, no scripts.

License

MIT. Third-party code is listed in THIRD_PARTY_NOTICES.md.

About

Open-source AI interface builder: drag components onto a canvas, ask an AI to build pages, export HTML.

Topics

Resources

Stars

3 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages