AI News

Google Releases Generative UI Protocol A2UI 1.0, ByteDance Lynx Already Adapted in Advance

Phoenix TechnologyUpdated Nov 12 07:55

AI summary

Google announced on November 11 local time that version 1.0 of A2UI, an open protocol for generative UI, is officially live, and the official documentation now lists v1.0 as the current version, enabling AI agents to respond with interactive native interfaces rather than plain text. Version 1.0 adds bidirectional function calling, accessibility support, and metadata extensions; agents only describe the content to be displayed, and each app draws it with native components, with sensitive values never entering the model. The protocol is open-sourced under the Apache 2.0 license, created by Google with contributions from CopilotKit and the open source community. Google provides open source renderers for React, Angular, Lit, Flutter, SwiftUI, Jetpack Compose, and others, and the agent SDK adds TypeScript and Dart alongside Python; Atlassian is bringing A2UI into Jira, while ByteDance Lynx currently supports A2UI 0.9 and is working with the A2UI team to adapt to 1.0.

Why read this

AI score 69

Google has replaced agents' text answers with native interactive interfaces; the article lists renderers, security design, and real-world cases such as Jira, which can be used to judge how such protocols are actually integrated.

Topics

Tags

Body · AI translation

Google's new AI protocol officially released, ByteDance had already laid groundwork Phoenix Tech Download the app for exclusive early access November 12, 2026 06:25:49 From Beijing

Phoenix Tech News, November 12 Beijing time: Google announced on November 11 local time that version 1.0 of its generative UI open protocol A2UI is officially live, and the official documentation now lists v1.0 as the current version. Google said in its official blog that this is the protocol's first stable release, allowing AI agents to respond not only with text but with interactive native interfaces, such as confirmation forms or charts. The blog compared the current interaction between users and agents to computers still being in the terminal era: capable enough, but almost entirely text-based.

According to the official blog, version 1.0 adds bidirectional function calling, allowing flight or reservation selectors to check availability on the spot; it also adds accessibility support and extensions for attaching metadata. On security design, agents only describe what is to be displayed, and each app draws the interface with its own native components; data and subcomponents are bound on the server side, and sensitive values do not enter the model. The A2UI official website states that the protocol is open source under the Apache 2.0 license, created by Google with contributions from CopilotKit and the open source community.

On renderers, the official blog lists open source renderers for major platforms: React, Angular, and Lit for the web; Flutter covering mobile, web, and desktop; SwiftUI for iOS; and Jetpack Compose for Android. The Jetpack Compose renderer is currently in alpha, and the official statement says its 1.0 support is coming soon. The Web and Flutter renderers can also work within MCP Apps. On agent SDKs, the official release adds TypeScript and Dart alongside the existing Python, which the official statement says can be used to generate skills, translate messages, and correct errors. Google Cloud documentation shows that enterprises can already register their own A2UI agents in Gemini Enterprise, which currently supports protocol 0.9.1.

Comparison of the same Alation feed rendered in Claude (left) and in the Alation data product app (right)

On use cases, the official blog says Atlassian is bringing A2UI into Jira: teams can assign work items to AI agents, and when an agent needs additional details or a decision, it asks the team questions within the work item via interactive cards, and the team answers directly in Jira. Alation's intelligent operating system (AIOS) uses A2UI to build metric cards and chart feeds, and each chart can show the SQL query and data behind it. When a user asks Claude about a profitable product feed, the feed can render within the conversation; the same feed can also be viewed in Alation with a nearly identical layout. In June 2026, ByteDance released an A2UI renderer for its cross-platform framework Lynx, and its Lynx GenUI Playground can natively render A2UI messages into interfaces such as a two-day Kyoto itinerary. The official page states that Lynx currently supports A2UI 0.9 and is working with the A2UI team to adapt to 1.0 and A2UI Express.

Official page — in the Lynx GenUI Playground, an A2UI message (left) is natively rendered as a Kyoto itinerary planning interface (right)

Inside Google, Gemini Notebook has already adopted A2UI: when a user asks for slides, the agent first pops up options asking whether they want slides for a presentation or a detailed document, and which topics to focus on; it then presents a review card where users can modify the format, topics, and visual style before generating. The official page says the same cards are also used for Video Overviews on web and mobile. Google says A2UI interfaces can also appear in AI assistants that support MCP Apps, such as Claude and Gemini; the official documentation and quick start address is a2ui.org. (Editor/Qiongqi)

For more first-hand news, welcome to download the Phoenix News app and subscribe to Phoenix Tech. For in-depth reporting, search "Phoenix Tech" on WeChat.

Source

Phoenix Technology