Generative UI: Specs, Patterns, and the Protocols Behind Them (MCP Apps, A2UI, AG-UI)

​Generative UI is quickly becoming one of the most important shifts in how agent-powered applications are built. Instead of agents returning only text, they can now drive real UI: structured components, dynamic layouts, interactive elements, and stateful experiences that live directly inside your product.

​In this session, we’ll walk through the different Generative UI specs and patterns emerging across the ecosystem, and, more importantly, the protocols that make them possible.

​We’ll cover:

• ​What Generative UI actually is (and what it’s not)

• ​Common Generative UI patterns developers are using today

• ​How specs like A2UI and MCP-UI define what an agent can render

• ​How AG-UI acts as the real-time protocol that connects agents, state, and UI

• ​How these pieces fit together rather than compete

• ​Practical examples of how developers can implement these patterns in real apps using CopilotKit

​This isn’t a sales pitch or a high-level theory talk. The goal is to give builders a clear mental model for the Generative UI landscape, explain when and why you’d use each approach, and show how to move beyond chat-only experiences toward fully interactive, agent-native products.

​If you’re building AI features and wondering how to turn agent output into real UX, we welcome you to attend this session.