thesysdev / openui

OpenUI — The Open Standard for Generative UI OpenUI is a full-stack Generative UI framework — a compact streaming-first language, a React runtime with built-in component libraries, and ready-to-use chat interfaces — that is up to 67% more token-efficient than JSON. --- Docs · Playground · Sample Chat App · Discord · Contributing · Code of Conduct · Security · License --- What is OpenUI At the center of OpenUI is **OpenUI Lang**: a compact, streaming-first language for model-generated UI. Instead of treating model output as only text, OpenUI lets you define components, generate prompt instructions from that component library, and render structured UI as the model streams. **Core capabilities:** • **OpenUI Lang** — A compact language for structured UI generation designed for streaming output. • **Built-in component libraries** — Charts, forms, tables, layouts, and more — ready to use or extend. • **Prompt generation from your component library** — Generate model instructions directly from the components you allow. • **Streaming renderer** — Parse and render model output progressively in React as tokens arrive. • **Chat and app surfaces** - Use the same foundation for assistants, copilots, and broader interactive product flows. Quick Start This is the fastest way to start with OpenUI. The scaffolded app gives you an end-to-end starting point with streaming, built-in UI, and OpenUI Lang support. What this gives you: • **OpenUI Lang support** - Start with structured UI generation built into the app flow. • **Library-driven prompts** - Generate instructions from your allowed component set. • **Streaming support** - Update the UI progressively as output arrives. • **Working app foundation** - Start from a ready-to-run example instead of wiring everything manually. How it works Your components define what the model can generate. • Define or reuse a component library. • Generate a system prompt from that library. • Send that prompt to your model. • Stream OpenUI Lang output back to the client. • Render the output progressively with Renderer. Try it yourself in the Playground — generate UI live with the default component library. Packages | Package | Description | | :--- | :--- | | | Core runtime — component definitions, parser, renderer, prompt generation | | | Headless chat state, streaming adapters, message format converters | | | Prebuilt chat layouts and two built-in component libraries | | | CLI for scaffolding apps and generating system prompts | Why OpenUI Lang OpenUI Lang is designed for model-generated UI that needs to be both structured and streamable. • **Streaming output** — Emit UI incrementally as tokens arrive. • **Token efficiency** — Up to 67% fewer tokens than equivalent JSON (see benchmarks). • **Controlled rendering** — Restrict output to the components you define and register. • **Typed component contracts** — Define component props and structure up front with Zod schemas. Token efficiency benchmarks Measured with (GPT-5 encoder). OpenUI Lang vs two JSON-based streaming formats across seven UI scenarios: | Scenario | Vercel JSON-Render | Thesys C1 JSON | OpenUI Lang | vs Vercel | vs C1 | | ------------------ | -----------------: | -------------: | ----------: | ---------: | ---------: | | simple-table | 340 | 357 | 148 | -56.5% | -58.5% | | chart-with-data | 520 | 516 | 231 | -55.6% | -55.2% | | contact-form | 893 | 849 | 294 | -67.1% | -65.4% | | dashboard | 2247 | 2261 | 1226 | -45.4% | -45.8% | | pricing-page | 2487 | 2379 | 1195 | -52.0% | -49.8% | | settings-panel | 1244 | 1205 | 540 | -56.6% | -55.2% | | e-commerce-product | 2449 | 2381 | 1166 | -52.4% | -51.0% | | **TOTAL** | **10180** | **9948** | **4800** | **-52.8%** | **-51.7%** | Full methodology and reproduction steps in . Documentation Detailed documentation is available at openui.com. Repository structure Good places to start: • openui.com for the full docs • for a working app • if you want to contribute Community • Discord — Ask questions, share what you're building • GitHub Issues — Report bugs or request features Contributing Contributions are welcome. See for contribution guidelines and ways to get involved. License This project is available under the terms described in .