Google Open-Sources A2UI Protocol to Enable AI Agents to Generate Dynamic User Interfaces

Quick Facts

• Google released A2UI as an open-source protocol in December 2025 to enable AI agents to generate cross-platform user interfaces
• The protocol uses declarative JSON formats and component whitelisting to prevent security risks from malicious agents
• Major implementations include Google’s Opal platform and Gemini Enterprise, with Flutter GenUI SDK providing native rendering support

Google has open-sourced A2UI (Agent-to-User Interface), a new protocol that enables AI agents to generate dynamic user interfaces across platforms while maintaining security. The company released the protocol in December 2025 to address the growing need for AI agents that can do more than just provide text responses.

A2UI solves the challenge of how AI agents can safely send rich UI components across trust boundaries. The protocol treats all external agents as potentially malicious and operates on three core security principles: declarative data format using pure JSON with no executable code, component catalog whitelisting that limits agents to pre-approved components, and schema validation that checks every message against JSON schemas.

‘Our developers choose Flutter because it lets them quickly create expressive, brand-rich, custom design systems that feel great on every platform,’ said Vijay Menon, Engineering Director at Dart. ‘A2UI was a great fit for Flutter’s GenUI SDK because it ensures that every user, on every platform, gets a high quality native feeling experience.’

The protocol is already powering production systems at Google. The Opal platform uses A2UI to enable hundreds of thousands of users to build AI mini-apps using natural language without coding. Gemini Enterprise is integrating A2UI to allow enterprise agents to render interactive UIs within host applications.

‘Our customers need their agents to do more than just answer questions; they need them to guide employees through complex workflows,’ said Fred Jabbour, Product Manager at Gemini Enterprise. ‘A2UI will allow developers building on Gemini Enterprise to have their agents generate the dynamic, custom UIs needed for any task, from data entry forms to approval dashboards.’

Performance testing shows strong results. MoltBot’s implementation achieved sub-100ms render times for over 1,000 dynamic UI components on Apple Silicon Macs. Security testing blocked 100% of 500+ injection attack patterns through the component-based allowlist approach.

The launch comes as enterprise AI adoption accelerates. PwC’s May 2025 survey found 88% of senior executives plan to increase AI budgets in the next 12 months, with 79% reporting AI agents are already being adopted in their companies. Industry analysts forecast the AI market will grow at a 35.9% compound annual growth rate through 2030.

A2UI takes a ‘native-first’ approach that differs from web-centric competitors. The protocol works alongside other emerging standards rather than competing with them. CopilotKit and AG-UI both support A2UI as complementary layers in the AI agent technology stack.

Google currently has A2UI at version 0.8 in public preview, with specifications and implementations that remain functional but evolving. The company opened the project to gather feedback and contributions from the developer community.

Read more: Dynamic UI for dynamic AI: Inside the emerging A2UI model