Hands-on Kube-Agents: Building an autonomous, intent-driven Kubernetes platform

Deploy and run Kube-Agents in this hands-on workshop. You'll install the operator, secure the Platform Agent, run proactive fleet audits, and safely resolve cluster incidents through review-ready GitOps pull requests.

As Kubernetes fleets grow, platform and SRE teams face mounting toil from manual debugging, config drift, and triage. Generic AI can parse logs, but it lacks live cluster awareness and deterministic guardrails. Kube-Agents (github.com/gke-labs/kube-agents) is an open-source agentic harness that continuously compares live cluster state against your declared operational intent. It inspects clusters read-only and makes changes only through verified GitOps, so it can surface silent degradation before it becomes an outage.

In this workshop, you'll deploy the operator, configure Workload Identity, trigger proactive health patrols, and fix live cluster incidents with pull requests validated in sandboxes before merge. You'll leave with a working setup and a guardrail pattern you can apply to any agent touching production: read-only access, PR-only writes, and sandbox validation.

Prerequisites: familiarity with kubectl and GitOps workflows. A GKE cluster is recommended.