A2UI in Google Chat and Gemini Enterprise: Product Boundaries Explained

A2UI is a protocol, but deployment succeeds or fails at a product boundary. Google Chat and Gemini Enterprise both expose A2UI workflows, yet their documented paths are not interchangeable.

Gemini Enterprise

Gemini Enterprise documents registration of A2UI-capable agents, including A2A-based agents, and currently identifies a v0.8 compatibility boundary for that route. A team using v0.9.1 elsewhere therefore needs a deliberate adapter or a separately versioned agent endpoint.

Managed ADK-based experiences and registered A2A agents also differ in identity, deployment, discovery, and lifecycle. Record which route is in use before copying configuration or message examples.

Google Chat

Google Chat documents an A2UI quickstart using ADK and Vertex AI Agent Engine. The feature is described as an early public preview, so eligibility, APIs, behavior, and support guarantees can change. Preview status should be visible in the architecture decision and release plan.

Chat also imposes its own interaction surface, authentication, tenancy, and deployment model. A renderer sample that works in a standalone web application is not evidence that it will work unchanged inside Chat.

Compatibility checklist

For either product, verify the accepted A2UI version, supported catalogs, agent runtime, region, identity model, preview terms, quotas, logging behavior, and fallback experience. Test unsupported components, long streams, user cancellation, expired authorization, duplicate actions, and partial failures.

Do not send private prompts, credentials, or unnecessary personal data through the UI model. High-impact actions still require server-side authorization and explicit user confirmation. The complete product matrix lives in our Google integration guide; protocol differences are tracked in version status.