A2UI

A2UI is a framework-agnostic standard for declaring AI-generated UI intent so agents can drive client interfaces using the client’s existing component catalog.

Key points

• Google frames A2UI v0.9 as a production-oriented answer to generative UI’s separation-of-concerns problem: agents declare UI intent, while clients render through their own design systems and frameworks ^[src-038].
• A2UI is meant to work across web, mobile, and other user surfaces, supporting local or remote agents that communicate with client applications through a common UI language ^[src-038].
• Version 0.9 renames the optional component set from “Standard” to “Basic” to clarify that frontend teams can bring their own components rather than adopt a new component system ^[src-038].
• The release adds a shared web-core library, official React renderer, updated Flutter/Lit/Angular/React renderers, and a space for community renderers ^[src-038].
• The Agent SDK helps agents generate A2UI JSON, manage spec versions, produce system prompts, parse and repair model output, validate against catalogs, and stream UI parts ^[src-038].
• A2UI supports simplified transports, including MCP, WebSockets, REST, AG-UI, and [[agent-to-agent-protocol|A2A]] ^[src-038].
• Production features include version negotiation, dynamic catalogs, resilient streaming, client-defined functions, client-to-server data syncing, improved errors, and a modular schema ^[src-038].

Related entities

• Google Developers Blog

Related concepts

• Generative UI
• Component Catalog Driven UI
• Agent-to-Agent Protocol
• Model Context Protocol (MCP)
• Agentic AI
• Visual Prototyping with AI

Source references

• ^[src-038] Google A2UI Team — “A2UI v0.9: The New Standard for Portable, Framework-Agnostic Generative UI” (2026-04-17)