Building an agent that works once is easy. Building one that survives crashes, retries and long tasks in production is the hard part. In this session I'll show how to wrap your agents in a durable harness using Amazon Bedrock AgentCore for the agent runtime and memory, and Temporal for the durability layer that keeps state and retries alive through any failure. We'll look at where agents break and how to build ones you can actually trust to run on their own.