LLMs are no longer just conversational engines, they’ve become the foundation for autonomous agents that can act, reason over tools and APIs, and work alongside other agents. But despite this progress, many AI agent frameworks still aren’t built to run natively on Kubernetes.

In this talk , we explore how to build declarative AI agents the cloud-native way, treating agents as first-class platform resources on Kubernetes. You’ll see how kagent, a CNCF sandbox project, enables platform teams to define, run, and govern agents using familiar Kubernetes patterns bringing structure, observability, and scalability to agent-based systems.