A2UI (Agent-to-User Interface) is an open-source protocol and renderer ecosystem that enables AI agents to generate native UI across any platform. Agents produce declarative JSON describing what to show; your client renders it using its own trusted components — no code execution, no iframes, no security risk.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/a2ui-project/a2ui/llms.txt
Use this file to discover all available pages before exploring further.
Quickstart
Run a Gemini-powered agent with A2UI rendering in under 5 minutes.
Core Concepts
Understand surfaces, components, data binding, and the message lifecycle.
Client Setup
Integrate A2UI into React, Angular, Lit, or Flutter.
Agent Development
Build agents that generate A2UI JSON with Gemini ADK or any LLM.
Why A2UI
Text-only agent interactions require multiple round-trips. A form that collects a restaurant booking takes five exchanges in chat — one well-designed UI collapses it to a single interaction. A2UI gives agents the ability to generate that UI without the security risk of running arbitrary code on your client.Security First
Declarative JSON, not executable code. Your client’s component catalog is the allowlist — agents can only request pre-approved components.
Framework Agnostic
One A2UI payload renders on React, Angular, Lit, and Flutter. The same agent response works everywhere.
LLM Friendly
Flat adjacency-list structure is easy for LLMs to generate incrementally and stream progressively.
How It Works
Agent generates UI
Your AI agent (using Gemini, GPT-4, or any model) produces a sequence of A2UI JSON messages describing the component tree and data model.
Messages travel over transport
The JSON stream is sent to your client via A2A Protocol, AG-UI (CopilotKit), WebSockets, or SSE.
Renderer parses and resolves
Your client’s A2UI renderer (React, Angular, Lit, or Flutter) processes the messages and looks up each component in your trusted catalog.
Choose Your Path
I'm building the frontend
Set up a renderer and register your component catalog.
I'm building the agent
Generate A2UI JSON from an LLM and stream it to your client.
I'm using AG-UI / CopilotKit
Enable A2UI in minutes with any CopilotKit-supported agent framework.
I'm building a renderer
Implement A2UI for a new platform or UI framework.
Current Status
A2UI is in early-stage public preview. The current stable release is v0.9.1. The v1.0 specification is a release candidate.v0.9.1 Specification
Current stable protocol — recommended for all new projects.
Roadmap
Upcoming renderers, transports, and agent framework integrations.