Google announced A2UI (Agent to UI) in February 2026, a declarative UI protocol designed for agent-driven interfaces that prioritizes security over executable code approaches. The open-source protocol enables AI agents to generate JSON payloads describing UI components while client applications maintain a catalog of pre-approved, trusted components, preventing arbitrary code execution vulnerabilities.
Security-First Architecture Prevents Arbitrary Code Execution
Unlike traditional generative UI approaches that allow agents to execute arbitrary code, A2UI implements a catalog-based security model. Agents can only request rendering of components from a pre-approved catalog maintained by the client application, such as Card, Button, or TextField components. This architecture ensures that AI agents cannot inject malicious code while still enabling dynamic interface generation.
Cross-Platform Rendering With Native Performance
The protocol represents UI as a flat list of components with identifier references, making it easier for language models to generate or update interfaces incrementally while supporting streaming updates. A single A2UI JSON payload can be rendered across multiple platforms built on different frameworks, with each client inheriting its native styling, accessibility features, and performance characteristics. Client libraries currently exist for Flutter, Web Components, and Angular.
Google Teams Already Integrating A2UI Into Production
The protocol is Apache 2.0 licensed and available on GitHub at google/A2UI. Google's Opal team has been a core contributor, using A2UI to rapidly prototype and integrate into their core app building flow. The GenUI SDK uses A2UI as the UI declaration format between remote server-side agents and applications, demonstrating production readiness.
Competing With Emerging Agent UI Standards
A2UI represents a major 2026 development in standardizing how AI agents communicate interactive user interfaces across platforms. The protocol competes with other emerging standards like MCP Apps and AG-UI protocol as the industry converges on standardized approaches for agent-driven interface generation.
Key Takeaways
- A2UI uses a security-first design where agents generate JSON payloads requesting pre-approved UI components from a client-maintained catalog, preventing arbitrary code execution
- The same A2UI JSON payload can render on multiple platforms (Flutter, Web Components, Angular) while inheriting each app's native styling and performance
- Google's Opal team and GenUI SDK are already using A2UI in production workflows for rapid prototyping and agent-driven app building
- The protocol is Apache 2.0 licensed and available on GitHub at google/A2UI with client libraries for major frameworks
- A2UI competes with emerging standards like MCP Apps and AG-UI protocol as the industry standardizes agent-driven interface protocols