What is A2UI?

A2UI is an open protocol by Google that enables AI agents to generate rich, interactive user interfaces without executing arbitrary code. Instead of relying on risky script injection or custom rendering logic, agents send declarative JSON payloads that client applications render natively. This approach works across Flutter, Web, and Mobile platforms, making it framework-agnostic and inherently secure. By separating UI generation from execution, A2UI gives developers a standardized way to let agents build dynamic interfaces that users can interact with directly.

Who it's for

• Agent developers who want their AI systems to produce interactive UIs without writing platform-specific rendering code or exposing users to code execution risks.
• Frontend engineers building applications that need to display agent-generated content—forms, dashboards, charts, or maps—using a consistent, secure protocol across Web, Flutter, and mobile clients.
• Product teams exploring agent-driven experiences who need a production-ready standard that works with existing agent frameworks like MCP and A2A, without reinventing the wheel.

Key features

Declarative JSON protocol

Agents describe UIs using structured JSON instead of executable scripts. This eliminates entire classes of security vulnerabilities while keeping the interface definition clean, parseable, and easy to validate. The protocol is designed specifically for LLM output, so agents can generate valid UI descriptions reliably.

Native rendering across platforms

A2UI clients render the declarative JSON natively on Flutter, Web, and mobile platforms. There's no need for iframes, sandboxed JavaScript, or custom WebView hacks. The same agent output produces a consistent, performant UI regardless of the user's device or framework.

Built-in data binding and actions

Components can bind to live data sources and trigger actions when users interact with them. This means agents can build dashboards that update in real time, forms that submit data back to the agent, or interactive maps that respond to clicks—all through the same declarative JSON structure.

Extensible component catalogs

Developers can define custom component catalogs beyond the built-in gallery. If your application needs specialized charts, maps, or domain-specific widgets, you can register them with A2UI and let agents use them just like any standard component.

What stands out

"A2UI lets agents build UIs without writing code—just declarative JSON that clients render safely."

This is the core innovation: instead of forcing agents to generate HTML, JavaScript, or platform-specific code, A2UI gives them a structured language for describing interfaces. The result is a secure, predictable way to let AI drive user interactions. Combined with support for MCP and A2A integrations, A2UI fits naturally into existing agent ecosystems without requiring a complete rewrite of your stack.

Worth checking out if…

You're building agent-powered applications and want a secure, standardized way to generate interactive UIs without executing arbitrary code. A2UI is especially relevant if you work with Flutter, Web, or mobile clients and need a protocol that works across all of them. It's also a strong fit if you're already using MCP or A2A and want to add rich UI capabilities to your agents without introducing security risks or platform lock-in.