# A2UI

> A protocol enabling AI agents to generate rich, interactive user interfaces that render natively across web, mobile, and desktop without executing arbitrary code.

A2UI is an open-source protocol created by Google that enables AI agents to safely generate rich, interactive user interfaces across trust boundaries. Instead of text-only responses or risky code execution, A2UI allows agents to send declarative component descriptions that clients render using their own native widgets, essentially providing a universal UI language for AI agents.

The protocol solves a critical challenge in agent development: how can AI agents safely send rich UIs without security risks? A2UI addresses this through a secure, declarative data format rather than executable code.

- **Secure by Design** - Uses a declarative data format instead of executable code, allowing agents to only use pre-approved components from your catalog, preventing UI injection attacks.

- **LLM-Friendly Architecture** - Features a flat, streaming JSON structure designed for easy generation, enabling LLMs to build UIs incrementally without requiring perfect JSON output in a single response.

- **Framework-Agnostic Rendering** - One agent response works everywhere, allowing the same UI to render on Angular, Flutter, React, or native mobile platforms with your own styled components.

- **Progressive Rendering** - Supports streaming UI updates as they're generated, so users see the interface building in real-time instead of waiting for complete responses.

- **A2A Extension Support** - Includes an extension for Agent-to-Agent communication, enabling complex multi-agent workflows with UI generation capabilities.

- **Custom Components** - Allows developers to create custom components like interactive charts and maps that agents can choose to use when responding to user queries.

- **A2UI Composer** - Provides a widget builder tool (via CopilotKit) for experimenting with and creating A2UI interfaces.

To get started, developers can run the quickstart guide featuring a restaurant finder demo powered by Gemini agents. The documentation covers core concepts including surfaces, components, data binding, and the adjacency list model. Developer guides help integrate A2UI renderers into applications or build agents that generate UIs. The protocol is currently at v0.8 (Public Preview) and is Apache 2.0 licensed, with active development on GitHub and contributions from CopilotKit and the open source community.

## Features
- Declarative UI generation for AI agents
- Secure by design - no executable code
- LLM-friendly streaming JSON structure
- Framework-agnostic rendering
- Progressive real-time rendering
- Custom component support
- A2A extension for agent-to-agent communication
- Native rendering across web, mobile, and desktop
- Data binding capabilities
- Theming and styling support

## Integrations
Angular, Flutter, React, Gemini, CopilotKit, Google Maps

## Platforms
WEB, DEVELOPER_SDK, API

## Pricing
Open Source

## Version
0.8

## Links
- Website: https://a2ui.org
- Documentation: https://a2ui.org/quickstart/
- Repository: https://github.com/google/A2UI
- EveryDev.ai: https://www.everydev.ai/tools/a2ui