OpenUI

OpenUI is an open source toolkit for making AI agents respond with a user interface. It presents itself as an open standard for generative UI and includes a playground, docs, and examples for working with OpenUI Lang rather than raw code or JSON.

Its workflow is described in four steps: a user defines a library with defineComponent and createLibrary, OpenUI generates a system prompt from that library, the model returns token-efficient, line-oriented OpenUI Lang, and a renderer parses the output and renders the interface in real time. The site also says it works for dashboards, support tools, developer tools, and project management, and names analytics, CRM and sales, customer support, DevTools, and project management as example use cases.

The product emphasizes live data, with interfaces able to query tools and MCP servers at runtime so the data stays current. It also describes cross-platform rendering to React, React Native, Vue, and more; streaming-first rendering that updates progressively as the model responds; interactive state, inputs, and actions; and a safety model in which the model only composes registered components and does not run arbitrary code. OpenUI also supports bringing existing components and design systems, and says it works with any LLM, any UI library, and agent frameworks including Vercel AI SDK, LangChain, CrewAI, OpenAI Agents SDK, and Anthropic Agents SDK.

OpenUI Cloud is introduced as a production-ready offering with tested, responsive components, editable artifact generation for slides, reports, dashboards, and pages, output validation, model and provider resilience, observability and an audit trail, pre-tested components, built-in accessibility, template support, editable text and charts, editable layouts, export as PPTX and PDF, version pinning, rollbacks, and provider fallbacks. The page attributes the project to Thesys Inc. and includes a command-line install path via pnpx, bunx, yarn dlx, or npx for @openuidev/cli@latest create.