Google Released A2UI v0.9 — A Standard for AI-Generated Interfaces

Google's A2UI team shipped version 0.9 of A2UI this week — a framework-agnostic standard that allows AI agents to generate user interface components in real time, using whatever design system and component library an organization already has in place. It is an open standard, not a proprietary platform, and it works across web, mobile, and any other surface where users interact with software.

The core premise: AI agents should be able to drive your front end without requiring you to rebuild it. Your components stay yours. The agent learns to speak your UI language.

What Generative UI Actually Means

Static interfaces present fixed options in fixed layouts. A user navigates menus, fills out forms, and interacts with components designed in advance for anticipated use cases. Generative UI inverts that model: the interface itself is generated in real time, shaped by the specific context of the user's current interaction.

A health companion app built on A2UI, for example, doesn't present a generic dashboard. It generates the specific widgets relevant to what the user is asking about right now — lab results, vaccine expiration dates, clinic locations — surfaced dynamically rather than buried in a static menu structure. A financial planning app generates sliders, charts, and multi-select components specific to a user's stated goal, rather than presenting a one-size-fits-all interface.

The user experience difference is significant: instead of navigating a product to find what they need, the product reorganizes itself around it. The AI does the navigation.

What A2UI v0.9 Actually Ships

The v0.9 release focuses on three areas: developer experience, framework compatibility, and production-grade reliability.

On developer experience: the release introduces an Agent SDK that simplifies building the agent side of a generative UI. Integration is a five-step process — define your component catalog, initialize the schema manager, generate the system prompt, initialize your AI agent with those instructions, and execute the streaming UI response. The SDK handles parsing, validation, and error correction of the AI's JSON output in real time, so components render as they are generated rather than waiting for a complete response.

On framework compatibility: A2UI v0.9 introduces a shared web-core library that simplifies browser UI rendering, and ships official support for React, Flutter, Lit, and Angular renderers, with a dedicated path for community-built renderers. The standard works over MCP, WebSockets, REST, and other transport protocols, including the newly launched Agent-to-Agent (A2A) 1.0 protocol.

On production reliability: the SDK supports version negotiation between agents and clients, dynamic catalog switching at runtime based on user permissions or device constraints, and resilient streaming that incrementally parses and repairs AI output — allowing partial UI rendering rather than all-or-nothing responses.

The component naming change from "Standard" to "Basic" reflects a deliberate philosophical clarification: A2UI's optional built-in components are a fallback, not the point. The point is that agents use your components.

The Ecosystem Building Around A2UI

A standard without ecosystem adoption is just a proposal. A2UI is accumulating meaningful integrations.

AG2, built by the creators of AutoGen, now supports native A2UI. Vercel launched a json-renderer that supports A2UI as a proof of concept — a potentially significant distribution path given Vercel's presence in the web development community. Oracle shipped Agent Spec with A2UI support as part of a layered architecture in which Agent Spec defines what runs, AG-UI handles interaction, and A2UI defines what the user touches. Each layer is independently swappable.

The architectural logic of that Oracle stack is worth understanding: it describes a future in which the agent layer, the communication layer, and the interface layer are decoupled. Organizations can upgrade or replace any one layer without rebuilding the others. That modularity is what makes a standard approach more durable than proprietary integrations.

Real Implementations: Health and Finance

Two production-adjacent implementations illustrate what A2UI enables beyond developer tooling.

The GenUI Personal Health Companion, developed by Rebel App Studio and Codemate, replaces static health dashboards with a chat-driven interface that generates UI widgets on the fly based on a user's actual health data — bridging fragmented medical records and wearable telemetry in a single conversational interface. The app is open source.

The Life Goal Simulator, built by Very Good Ventures for the financial services sector, uses Gemini and Flutter's GenUI SDK to generate native-feeling interfaces based on a user's specific financial goal. Select a persona and an objective — saving for retirement, buying a home — and the agent generates the relevant interactive components: sliders, charts, multi-selects — specific to that goal rather than generic across all users.

Both implementations share the same underlying pattern: the interface is a response to the user's intent, not a fixed structure the user has to navigate.

What This Means for Product and Marketing Teams

For product teams, A2UI represents a shift in how interface design relates to user experience. If AI agents can generate contextually appropriate UI in real time, the design work moves upstream — from designing specific screens to designing component systems that agents can compose intelligently. The product surface becomes dynamic rather than fixed.

For marketing teams, the implications follow from those of the product. Personalization at the interface level — not just content personalization, but structural personalization where the UI itself adapts to the user — changes what a high-performing digital experience looks like. The conversion optimization, the journey mapping, the content strategy built around fixed page structures: all of it becomes more fluid as generative UI matures.

This is not yet the default mode for most digital products. A2UI v0.9 is a standard in active development, not a production norm. But the direction is clear, the ecosystem is building, and the organizations thinking now about how their component libraries, design systems, and AI agent strategies intersect will be better positioned when generative UI becomes a baseline expectation rather than a differentiator.

Understanding where AI infrastructure is heading — and what it means for how products and marketing experiences are built — is core to what our team at Winsome Marketing does with growth-focused clients. If you want to think through what the generative UI shift means for your product or marketing strategy, let's connect.