Today’s interfaces often expose agents as simple chats or side panels, while real products are organized around clear domains and responsibilities. This session explores how to use A2UI to deliver complete agentic experiences as microfrontends, each aligned to a specific product context and owned by a dedicated team. Instead of sprinkling agent calls across many components, an agentic microfrontend becomes the primary interface for a given task space, with server-side agents describing UI updates via A2UI and the microfrontend rendering them using its own trusted design system.

The talk walks through the core patterns for this approach: structuring an agentic microfrontend around a domain boundary; using A2UI as the contract between agents and UI; handling streaming updates and state transitions over time; and integrating with existing navigation, auth, and observability. Particular emphasis is placed on production concerns: how to test agent-driven flows, how to constrain and validate A2UI payloads, how to instrument interactions end-to-end, and how to keep agent behavior evolvable without breaking the surrounding product.

Attendees gain clear mental models for where agentic microfrontends make sense, practical guidance on designing the A2UI contract and lifecycle, and concrete patterns they can apply when introducing agent-driven experiences into complex applications and organizations.