ClusterDoctor

Pick a cluster and a namespace. Cluster Doctor finds what is broken, names the root cause in one line and gives you the commands to fix it

Get Started

Steer to What Broke

A real scan of the kubeiq-demo namespace found 9 failing workloads. Pick one, or turn the wheel, and read its entry in the log

Log of kind-kubeiq-dev22 Oct 2026, 23:04:40

Bearing 000°·Confidence 94paymentDeployment in kubeiq-demo

Position: Deployment/kubeiq-demo/payment
Observed: Environment variable DATABASE_URL is missingAdd DATABASE_URL to the container env, usually from a Secret or ConfigMap
Action: kubectl set env deployment/payment DATABASE_URL=<database-url> -c app -n kubeiq-demo
Check: kubectl rollout status deployment/payment -n kubeiq-demo --timeout=120s
kubectl get pods -n kubeiq-demo -l app=payment
Affected: Deployment/kubeiq-demo/paymentPod/kubeiq-demo/payment-6858dfbd56-bv8fnPod/kubeiq-demo/payment-6858dfbd56-tj6s7
Next time: Validate required configuration at startup and in CI before deploying

How a Scan Runs

Every scan takes the same four steps, and none of them changes your cluster. The one on this page took 15 seconds and checked 11 pods, 9 workloads, 8 events and 13 logs

Read the Cluster

Pods, events, workloads, services, nodes and logs, read with the access your kubeconfig already has. Secret values are never read

Cluster Manager listing the pods of kubeiq-demo with their status and restarts

Name the Cause

Known failures are recognized in seconds and their symptoms grouped under one root cause: 126 restarts of payment become one missing variable

Cluster Doctor naming the root cause of payment: environment variable DATABASE_URL is missing

Dig Deeper

When the cause is unclear, an AI agent reads previous logs, ConfigMaps and related resources with read-only tools. Passwords and tokens are masked first

The investigation steps: evidence collected, problems found, and the agent checking orders and queue-worker

Fix and Check

Commands with your real names, filled from the cluster, and the command that proves the fix rolled out. Changes stay off until an admin turns them on

The fix command for payment and the commands that check the rollout

Known on Sight

The failures Cluster Doctor recognizes in seconds: CrashLoopBackOff, OOMKilled, ImagePullBackOff, Pending pods, Probe failures, Missing config, Service selectors, Init containers, Failed jobs, Stuck rollouts, Node pressure. Anything else goes to the AI agent, so an unhealthy pod is never reported as fine

Safe to Point at Production

Cluster Doctor looks and explains. Changes happen only when you run the commands yourself, or in Cluster Manager once an admin allows them

Read-Only: Cluster Doctor uses read-only access and never reads secret values, only their names and metadata
Private by Default: Passwords, tokens and keys in logs and config are masked before anything reaches a model
Your Model, or None: OpenAI, Anthropic, OpenRouter, Bedrock, Azure, a LiteLLM proxy or a local model. Without one, known failures are still found and fixed

Install KubeiQ once and Cluster Doctor is ready under the Products menuInstall