As AI agents take on increasingly autonomous tasks, developers need to solve for identity, memory, and coordination, especially in distributed systems. This session explores how to build agents that can act independently, maintain long-term state, and securely interact with external services.

We’ll walk through how agents can use the Model Context Protocol (MCP) to communicate with APIs and services, how to implement authentication and authorization flows using identity providers like Auth0, Stytch, or WorkOS, and how Durable Objects provide a persistent, low-latency memory layer for agents. You’ll also learn how agents can enter a hibernation state during inactivity, reducing resource usage while remaining responsive.