Connect a Kubernetes cluster

A lightweight in-cluster agent pushes workload metrics (CPU/memory usage per Deployment, StatefulSet, Node, ...) every few minutes. No inbound access to your cluster is ever required — the agent only makes outbound calls to PruneOps.

How it works

One outbound-only agent per cluster.

01
Push-based

The agent is stateless and pushes samples on its own schedule — a restart never loses data, it just resumes from current state.

02
Per-cluster token

Each cluster gets its own bearer token, scoped to this organization only. Revoke or regenerate it any time.

03
Same pipeline as AWS

Workload metrics feed the same hourly analysis and right-sizing recommendations as any other resource type.

Setup

  1. Go to Organization → Integrations → Kubernetes and register a cluster. This generates an agent token, shown once.
  2. Install the agent with Helm, using the token and cluster ID from the previous step: 
    helm upgrade --install pruneops-k8s-agent oci://registry-1.docker.io/pruneops/k8s-agent-chart \
  --version 0.2.0 \
  --set pruneops.token=<TOKEN> \
  --set pruneops.clusterId=<CLUSTER_ID> \
  --namespace pruneops --create-namespace
  3. Wait a few minutes, then check the cluster list — "Last seen" updates once the agent starts reporting.