Google A2UI 1.0, Express Method Cuts Input Tokens by 45.4%

·2026.11.13 11:30

Key point

Google released A2UI 1.0, reducing input tokens by 45.4% when generating agent UIs.

1 / 6

Details

  • The Google A2UI team released A2UI 1.0, the open protocol for generative UI, on November 11, 2026.
  • Using A2UI Express reduces input tokens by 45.4% and generated code tokens by 63.2% compared to direct JSON generation (based on Google's internal evaluation).
  • Support for bidirectional function calls (bidirectional RPC) and mixed multiple catalogs expands interaction between agents and renderers.
  • When integrated with MCP Apps, A2UI screens can be provided to host apps via a web renderer inside an iframe.

Technical Changes

  • Maintains a declarative data (JSON)-based UI description method; apps render only using a pre-registered native component catalog.
  • Introduces @ prefix reserved keys (@path, @call, etc.) and removes theme to separate layout and branding.
  • Adds WAI-ARIA accessibility attributes and defines the return structure for input validation results (ValidationResult).

Support Status and Roadmap

  • React, Angular, Lit, Flutter, and SwiftUI renderers support 1.0; Jetpack Compose is scheduled to transition to 1.0.
  • Provides Python, TypeScript, and Dart Agent SDKs; Macro functionality is supported exclusively in the Python SDK.
  • Distributed under the Apache License 2.0, allowing free use for research and commercial purposes.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.