Kubectl MCP Server by rohitg00: Setup for Claude, Cursor and Other Apps

• Lists Kubernetes resources of any kind

  Reads resources such as pods, deployments, services or custom resources, in one namespace or all of them. It returns what kubectl get would show.

  Try: “Show all pods in the kube-system namespace.”

  Read · kubectl_get

• Shows the full details and events of one resource

  Pulls the detailed description of a single resource, including its recent events. Useful when you need to know why something is not starting.

  Try: “Describe the nginx deployment in the default namespace.”

  Read · kubectl_describe

• Reads container logs from a pod

  Fetches the log output of a pod or one of its containers. You can ask for the earlier run of a restarted container.

  Try: “Show the last logs from the payments pod.”

  Read · kubectl_logs

• Shows CPU and memory use of pods and nodes

  Reports live resource use for pods or nodes. It needs the metrics server running in the cluster.

  Try: “Which pods use the most memory right now?”

  Read · kubectl_top

• Explains what a resource field means

  Looks up the built-in API documentation for a resource type or field, so the assistant can write a correct manifest.

  Try: “Explain the fields of a Deployment's rollout strategy.”

  Read · kubectl_explain

• Lists the resource types your cluster supports

  Returns every API resource type the cluster knows about, including those added by custom resource definitions.

  Try: “What resource types does this cluster support?”

  Read · get_api_resources

• Runs any read-only kubectl command

  Runs a kubectl command the other tools do not cover. The README says it only allows read verbs and blocks credential, connection and raw-output flags.

  Try: “Run kubectl auth can-i list secrets in the staging namespace.”

  Runs code · kubectl_generic

• Applies a YAML manifest to the cluster

  Creates or updates resources from a manifest you or the assistant provide. This changes your live cluster.

  Try: “Apply this deployment manifest to the staging namespace.”

  Write · kubectl_apply

• Edits fields of an existing resource

  Changes part of a resource with a strategic, merge or JSON patch, without sending the whole manifest.

  Try: “Patch the api deployment to use image tag 1.4.2.”

  Write · kubectl_patch

• Deletes a resource from the cluster

  Removes a named resource such as a pod, deployment or config map. The deletion happens in the live cluster.

  Try: “Delete the stuck test-runner pod in the ci namespace.”

  Delete · delete_resource

• Changes the replica count of a workload

  Scales a deployment, stateful set or similar workload up or down. Scaling to zero stops all its pods.

  Try: “Scale the frontend deployment to 5 replicas.”

  Write · kubectl_scale

• Restarts, pauses or rolls back a workload

  Runs rollout actions on a workload: restart, status, history, undo, pause and resume.

  Try: “Restart the checkout deployment and tell me when it is ready.”

  Write · kubectl_rollout

• Runs a command inside a running pod

  Executes any command you give it inside a pod's container, with a 60 second limit, and returns the output. It can run arbitrary commands in your cluster.

  Try: “Run 'ls /app' inside the api-7d9 pod.”

  Runs code · exec_in_pod

• Lists the clusters in your kubeconfig

  Shows the Kubernetes contexts in your kubeconfig and which one is active.

  Try: “Which clusters can you reach?”

  Files · list_contexts

• Checks that the cluster is reachable and healthy

  Runs a quick health check against the chosen cluster context.

  Try: “Is the production cluster healthy?”

  Read · health_check

• Works out why a pod keeps crashing

  Collects the pod's status, events and logs and returns an analysis of the likely cause.

  Try: “Why is the nginx-deployment pod crashing?”

  Read · diagnose_pod_crash

• Finds pods stuck in Pending and says why

  Looks for pods that are not scheduled and reports the reasons the cluster gives, such as missing resources or taints.

  Try: “Which pods are stuck pending, and why?”

  Read · detect_pending_pods

• Reports which operators and CLIs are installed

  Detects add-ons such as Flux, Argo CD, cert-manager, Kyverno and Istio plus local command-line tools, and says which optional toolsets would cover them.

  Try: “What Kubernetes add-ons are installed on this cluster?”

  Read · detect_ecosystem

• Lists Helm releases

  Shows the Helm releases installed in a namespace or across the cluster.

  Try: “List all Helm releases in the monitoring namespace.”

  Read · helm_list

• Shows the status of one Helm release

  Reports whether a release is deployed, failed or pending, with its last update.

  Try: “What is the status of the grafana release?”

  Read · helm_status

• Reads the values or manifest of a release

  Fetches stored information about a release, such as the values it was installed with.

  Try: “Show the values used for the prometheus release.”

  Read · helm_get

• Shows the revision history of a release

  Lists the past revisions of a release so you can see what changed and when.

  Try: “Show the revision history of the api release.”

  Read · helm_history

• Installs a Helm chart as a new release

  Installs a chart into a namespace with the values you give. This adds new resources to the cluster.

  Try: “Install the bitnami redis chart in the cache namespace with 3 replicas.”

  Write · install_helm_chart

• Upgrades an existing Helm release

  Moves a release to a newer chart or new values. The tool is marked destructive in its annotations.

  Try: “Upgrade the grafana release to the latest chart.”

  Write · upgrade_helm_chart

• Removes a Helm release and its resources

  Uninstalls a release, which removes the resources the chart created.

  Try: “Uninstall the old-staging release from the staging namespace.”

  Delete · uninstall_helm_chart

• Rolls a release back to an earlier revision

  Reverts a Helm release to a previous revision, replacing what is running now.

  Try: “Roll the checkout release back to revision 4.”

  Write · helm_rollback

Watch out for

• It can delete resources and uninstall releases

  delete_resource and uninstall_helm_chart act on the live cluster. Start the server with --read-only, or use a kubeconfig context that only has the rights you want to give the assistant. github.com

• exec_in_pod runs any command in your pods

  The assistant can run arbitrary commands inside any container it can reach. Read-only mode removes this tool. github.com

• It acts with your full kubeconfig access

  The server uses whichever context is active and can switch between your clusters if the multicluster toolset is on. Check the active context before you ask for changes. github.com

• Opt-in toolsets add up to 209 tools

  Only the core and helm toolsets are on by default. Extra toolsets such as backup, kind and rollouts add more delete and write tools, so turn on only the ones you need. github.com

• Secrets are redacted in results

  The README says Secret data, tokens and JWTs are replaced with [REDACTED] in tool output. HTTP mode binds to 127.0.0.1 by default. github.com

Security advisories