Google Releases A2UI 1.0, a Stable Generative UI Protocol

Google on 11 November 2026 released A2UI 1.0, which it calls “the first stable version of our open protocol for generative UI.” In a Google Developers Blog post the same day, A2UI product manager Vishal Motwani wrote that the protocol “lets AI agents answer with interactive screens, such as a form to confirm or a chart,” and that developers can use it now with “open-source renderer SDKs for the web, Android, and iOS and agent SDKs for Python, TypeScript, and Dart.”

The idea is deliberately narrow. “With A2UI, the agent describes what to show, and every app draws the UI with its own native components.” The host application receives A2UI messages and renders them from a registered design system, rather than executing model-written markup. Google says that split is what lets the same agent output appear in a company’s own app or in a third-party assistant that hosts MCP apps, “such as Gemini or Claude,” at “native speed.”

The A2UI v1.0 specification is marked “Status: Current.” The spec page still shows “Last Updated: Jun 8, 2026”; the date for this stable 1.0 release is the 11 November Developers Blog post. The open-source repository is a2ui-project/a2ui.

What 1.0 actually adds

Version 1.0, Motwani writes, “adds bidirectional function calls, so a flight or reservation picker can check availability on the spot.” It also “adds support for accessibility and extensions for any additional metadata.” Open-source renderers draw screens with React, Angular and Lit on the web, Flutter on mobile, web and desktop, SwiftUI on iOS, and Jetpack Compose on Android. One caveat sits in that list: the Compose renderer is “currently v0.9.1, and v1.0 coming soon.” The web and Flutter renderers, Google says, also work inside MCP Apps. New agent SDKs for TypeScript and Dart, “along with Python,” are meant to generate skills, translate messages and correct errors; in Python, “A2UI Macros” add programmatic component expansion and data binding.

Google also says “companies such as Atlassian, Alation, Bloomberg, and ByteDance are already building with it.” That list is Google’s claim. TSN has not independently verified deployments. The blog itself describes the first three named companies in more detail than a logo strip: Atlassian, it says, “is bringing A2UI to Jira,” so that an agent needing a missing detail or a decision can ask with an interactive card inside a work item. Alation is said to build feeds in its Intelligence Operating System with A2UI metric cards and charts, including a view that can render in Claude and again inside Alation. ByteDance, Google says, released an A2UI renderer for its Lynx framework in June 2026; Lynx “supports A2UI 0.9 today and is adapting to 1.0 with the A2UI team.” Bloomberg is named in the same sentence and is not given a worked example in the post. Those accounts remain Google’s.

Inside Google, the post says Gemini Notebook already answers a request for a slide deck with option and review cards built this way. That is also a first-party example, not a third-party audit.

A protocol like this sits next to two other Google surfaces TSN has already covered. The connected-apps platform preview is Google’s bid to put other companies’ apps inside Search and Gemini; A2UI is a way an agent in those hosts could answer with a native screen rather than a paragraph. And the Dart agent SDK arrives in a toolchain that already includes Genkit Dart 1.0, covered earlier in a development-tools roundup. The contrast with screenshot-to-UI code tools is sharper still: those products try to turn a picture of an interface into code, while A2UI starts from a structured description an agent emits, then asks the host app to render its own components.

A stable protocol is not a verified rollout

The useful news is that Google has frozen a 1.0 protocol, published renderers for the main client platforms, and shipped agent SDKs in Python, TypeScript and Dart. The useful limit is everything that is still a claim or a qualifier. Jetpack Compose is not yet on 1.0. The specification page still carries a June last-updated date. The Atlassian, Alation, Bloomberg and ByteDance names are Google’s “already building” list; the Lynx renderer, on Google’s own telling, is still being adapted from 0.9. None of that makes 1.0 a vapour release, but it does mean this should be reported as a stable open protocol with first-party examples, not as a verified multi-vendor production standard.

The facts to watch are a Compose 1.0 renderer, a Lynx build that actually speaks 1.0, and any of the named companies describing a shipped interface in their own channels. Until those appear, A2UI 1.0 is the first stable version of Google’s generative-UI protocol, available to developers now, with the adopter list still sitting on Google’s page.

Sources

1. Vishal Motwani, Google Developers Blog, “A2UI 1.0: Build delightful, fast, reliable, agentic experiences,” 11 November 2026. https://developers.googleblog.com/en/a2ui-10-build-delightful-fast-reliable-agentic-experiences/
2. A2UI project, A2UI specification v1.0 (Status: Current; page last-updated date 8 June 2026). https://a2ui.org/specification/v1.0-a2ui/

Related stories

Google Previews One Platform to Put Apps in Search and Gemini is the nearest recent Google surface for third-party apps inside Gemini, where an A2UI screen would be drawn by the host.

Also relevant:

• Screenshot to UI Code: The Tools That Actually Do It