Agents can now take on complex, multi-step work where they need to operate over longer timelines, across more systems, and with higher reliability requirements. Running these agents in production requires a dedicated harness and runtime to coordinate execution, manage context, and scale across many users and agents.
In this session, we’ll unpack the harness and runtime architecture required to run mission-critical agents in production. We’ll cover the components of the agent harness, from execution flow and subagent delegation to human-in-the-loop workflows and context management. We’ll then look at the runtime layer underneath, including durable execution, isolation, persistence, scheduling, authentication, and production-scale infrastructure.
