Google Releases A2UI v0.9 — An Open Standard for AI Agents That Generate UI From Your Design System
Google released A2UI v0.9, a framework-agnostic open standard that lets AI agents dynamically generate UI components from an application's existing design system catalog — without executing arbitrary code. Ships a Python Agent SDK, React renderer, and updated Flutter/Lit/Angular support, with MCP, WebSockets, and REST transport.
Original sourceGoogle released A2UI v0.9 on April 17, 2026 — an open-source, framework-agnostic standard that gives AI agents a safe, structured way to generate UI without executing arbitrary code. The core premise: agents declare UI intent using a declarative data format; the client renders using its own pre-approved components. No code injection. No runtime surprises.
v0.9 ships a shared web-core library, an official React renderer, and updated renderers for Flutter, Lit, and Angular. The new Python Agent SDK gives server-side AI agents a typed interface for declaring UI actions. Transport support covers MCP, WebSockets, and REST — making A2UI composable with most modern agent architectures.
The ecosystem is already forming. AG2 (the AutoGen team), Vercel's json-renderer project, and Oracle's Agent Spec all integrate with A2UI. Early production implementations are live from Rebel App Studio and Very Good Ventures, covering use cases from dynamic form generation to context-sensitive mobile UI.
The design philosophy is what distinguishes A2UI from pure vibe-coding tools like v0 or Bolt. Those tools generate complete UI components as code. A2UI assumes you already have a design system — the agent's job is to select and configure the right components, not generate new ones. This is the architecture that makes AI-generated UI practical in existing enterprise and consumer apps where introducing arbitrary new components is unacceptable.
Whether A2UI becomes a genuine standard or Google's preferred format depends on adoption by major frameworks and tooling vendors. The AutoGen and Vercel integrations are strong early signals. If the React/Flutter/Angular renderer ecosystem grows to include Vue, Svelte, and SwiftUI, this has a real shot at becoming load-bearing infrastructure for the next generation of AI-native apps.
Panel Takes
The Builder
Developer Perspective
“The 'declarative intent, client renders from your own components' model is exactly what I've been waiting for. I can let AI agents surface dynamic UI without worrying about component drift or design system violations. The Python Agent SDK and MCP support means it drops cleanly into my existing agent stack.”
The Skeptic
Reality Check
“Google has a long history of launching open standards that quietly get abandoned when internal priorities shift. A2UI has the right architecture but it needs ecosystem buy-in from Vue, Svelte, and iOS/SwiftUI to have staying power. The Vercel and AutoGen integrations are promising but this is still v0.9 — treat as promising beta infrastructure, not production bedrock.”
The Futurist
Big Picture
“A2UI could be the missing layer between AI reasoning and every existing UI framework. If it becomes the standard transport for agent-to-UI communication, every app's components become AI-accessible by default — which fundamentally changes how we think about UI development. This is the kind of unsexy infrastructure that powers the next platform shift.”