Skip to main content
OpenUI lets the LLM compose rich interfaces on the fly using OpenUI Lang and the standard OpenUI React component library. The server exposes a prompt tool so the model learns the syntax and available components, then a render tool streams OpenUI Lang into a Skybridge view that parses and displays it with OpenUI’s <Renderer />.

​
How it works

The MCP server exposes get-openui-prompt, which returns OpenUI Lang instructions generated from openuiLibrary.prompt() — component signatures, Skybridge-specific rules, and an example program. The model then calls render with an OpenUI Lang program. Unlike JSON Renderer or Manifest UI, the view does not wait for the finished tool result. useToolInfo exposes the partial input as the host streams ui/notifications/tool-input, so OpenUI’s <Renderer /> can paint the UI statement by statement while the model is still generating. OpenUI Lang is compact and stream-friendly: the first line defines root, while later lines can fill in forward references as the model generates details.

​
Skybridge APIs used

  • useToolInfo — read streaming tool input for progressive rendering

​
Third-party integrations