Skip to main content
Back

Flutter's GenUI SDK Lets AI Build Real Interactive App Screens at Runtime

Flutter's official GenUI SDK lets LLMs compose real native widgets at runtime using the A2UI protocol, replacing text walls with interactive UI

Repo, Full Access
  • Flutter's official GenUI SDK lets LLMs compose real native widgets at runtime using structured JSON, replacing text responses with interactive UI.
  • Built on A2UI, an open Agent-to-UI protocol co-developed with Google's A2A team; framework-agnostic and compatible with CopilotKit and AG UI from day one.
  • Works with any LLM that supports structured output; Gemini 2.5 Pro/Flash recommended, Firebase AI Logic for production, custom adapters for others.
  • Currently alpha (v0.10.2), requires Flutter 3.35.7+; API will change drastically, not production-ready for most teams yet.
  • Key roadmap items: streaming UI, full-screen composition, and Dart bytecode for true server-driven UI without app-store updates.
  • Install via pub.dev and explore examples in the GitHub repo.

Flutter just shipped an official answer to one of the most persistent frustrations in AI-powered apps: the chatbot that responds to "book me a flight" with a wall of text instead of, you know, a booking form. The GenUI SDK for Flutter lets an LLM compose real, interactive Flutter widgets at runtime, directly inside your existing app, without generating a single line of code.

This was the headline architectural story of the Flutter 3.44 release cycle, spotlighted at Google I/O 2026 in a session titled "Flutter + A2UI = GenUI" and widely described as the most architecturally significant Flutter announcement of the year. The repo has already crossed 1.7k stars, and the pub.dev package has racked up 25.7k downloads.

The core idea: UI as agent output

Instead of hard-coding the layout, you hand an AI agent a kit of approved UI components and a goal, and the agent composes a real Flutter interface at runtime based on what the user is actually trying to accomplish. This is a fundamentally different model from AI coding assistants that generate Dart source code. It uses a JSON-based format to compose UIs from your existing widget catalog, turning conversations or agent interactions into rich, intuitive experiences. State changes in the UI update a client-side data model, which is fed back to the agent, creating a powerful, high-bandwidth interaction loop.

The key distinction the team emphasizes: the UI is not generated as code. The LLM outputs structured JSON describing which widgets to render and with what data. Your app's widget catalog is the safety boundary. Each catalog item has a name, a data schema describing the data that widget expects, and a builder function that turns that data into a real Flutter widget. The catalog is the safety boundary: the agent can only compose from what you put in it.

GenUI features breakdown showing LLM-generated interactive UI with destination cards composed from widget catalog

A2UI: the protocol underneath

A2UI ("Agent-to-UI") is the protocol underneath GenUI: a declarative UI protocol for agent-driven interfaces. In plain terms, it is an open, JSON-based serialization format for UI that an LLM produces. Think of it as the wire format that travels between your agent and your Flutter renderer.