Connect a Kubernetes Cluster

6 min readUpdated June 2026

What you'll need

Connecting a cluster takes a few minutes. The collector runs inside your cluster and only needs outbound HTTPS connectivity to Cloptima — no inbound access and no changes to your application workloads.

kubectl access to the cluster you want to monitor
Helm available in your shell (the installer deploys a Helm release)
Outbound HTTPS from the cluster to Cloptima

Create the cluster in Cloptima

In Cloptima, start onboarding and choose Kubernetes (or Settings → Cloud Accounts → Add → Kubernetes). Give the cluster a display name (for example, prod-eks-cluster) and an optional region to help group clusters across environments. Then generate the installer script. The one-time installation token it produces is valid for 24 hours.

Run the installer in your cluster

The console shows a copy-ready command set. Run them in a shell with kubectl context pointed at the target cluster. Step 1 exports the one-time install token into your session; step 2 downloads and runs the installer with curl (a wget alternative is provided). The installer performs preflight checks, validates permissions, and installs the collector as a Helm release in the cloptima namespace.

Copy the exact export and installer commands from the onboarding screen — each script embeds a unique, time-limited agent token. If the token expires, generate a new installer script.

Verify the agent and finish

Once the agent checks in, Cloptima automatically activates the cluster account and advances onboarding — metrics and resources begin streaming shortly after. If it hasn't connected yet, confirm the Helm release is healthy and check the agent logs.

kubectl logs -n cloptima deploy/cloptima-k8s-agent

What you get

With the collector running, Cloptima attributes cluster spend down to the workload and surfaces optimization opportunities.

Cost by cluster, namespace, workload, and pod — direct and idle (unused-capacity) cost separated
Rightsizing recommendations with confidence and safety scores from real utilization
Idle and orphaned resource detection
PR cost impact comments when workload manifests change

Next steps

Map namespaces to teams with virtual tags for team-level showback, set budgets and anomaly alerts, and connect GitHub to get cost impact on pull requests before they merge.