New standards for AI agent UIs highlight key differences between Google's native-first A2UI and the web-centric MCP Apps favored by OpenAI.

The New Stack is a publication covering trends and technologies in cloud-native development, DevOps, and software delivery. Developers can learn about containerization, Kubernetes, and cloud computing, as well as explore topics such as microservices architecture, serverless computing, and continuous integration/continuous delivery (CI/CD) pipelines.

The New Stack

Two competing approaches are emerging for building AI agent user interfaces. Google's A2UI takes a native-first, cross-platform approach where agents send declarative UI blueprints that can be rendered using native widgets across web, mobile, and desktop. In contrast, OpenAI and Anthropic favor web-centric solutions through MCP Apps, which uses sandboxed iframes to render web-based widgets. A2UI supports Flutter, Web Components, and Angular, with streaming capabilities for progressive rendering. Meanwhile, MCP Apps standardizes interactive UIs within the Model Context Protocol ecosystem. The fragmented landscape includes multiple frameworks and protocols, creating complexity for developers building agent applications.

Agent UI Standards Multiply: MCP Apps and Google’s A2UI

What Is A2UI? A Cross-Platform, Native-First Approach

Understanding OpenAI’s Web-Centric UI Strategy

MCP Apps vs. A2UI: Key Differences Explained