OpenUI

OpenUI is an open standard and toolkit that lets AI applications respond with rich, dynamic user interfaces instead of plain text. Built by Thesys, it introduces OpenUI Lang — a token-efficient format that allows LLMs to generate and stream full component trees at runtime. It is up to 3x faster than JSON-based rendering approaches and uses up to 67.1% fewer tokens, making it well-suited for production AI applications.

• OpenUI Lang — A purpose-built language for generative UI; define your component library once and let LLMs respond with structured, validated UI trees.
• defineComponent & createLibrary — Register your own React (or other framework) components with Zod-validated props schemas so the renderer knows exactly how to instantiate them.
• Native Streaming — UI renders in real time as the LLM streams its response, enabling live dashboards, conversational apps, and more.
• Token Efficiency — Up to 67.1% fewer tokens compared to json-render, reducing cost and latency for every AI response.
• Cross-Platform Support — Works with React, React Native, Vue, and other frameworks; supports web apps, mobile apps, dashboards, and bottom trays.
• MCP & Tool Integration — Query MCP servers and tools at runtime to inject live data directly into generated UI.
• Safe by Default — No arbitrary code execution; the renderer validates all props against the registered schema before rendering.
• LLM Agnostic — Compatible with OpenAI, Anthropic, Gemini, Mistral, xAI, DeepSeek, and more.
• Framework Integrations — Works with Vercel AI SDK, LangChain, CrewAI, OpenAI Agents SDK, Anthropic Agents SDK, and popular UI libraries like ShadCN, Material Design, DaisyUI, and Base UI.
• CLI Scaffolding — Bootstrap a new project instantly with npx @openuidev/cli@latest create.
• Playground & Demo — Try OpenUI Lang live in the browser playground without any local setup.