Kubernetes MCP Server: Setup for Claude, Cursor and Other Apps

• Lists the pods in your cluster

  Lists pods across all namespaces. You can filter by label or field and page through long lists.

  Try: “Show me every pod that is not Running.”

  Read · pods_list

• Lists the pods in one namespace

  Lists pods in the namespace you name, with the same filters.

  Try: “List the pods in the payments namespace.”

  Read · pods_list_in_namespace

• Shows one pod in detail

  Returns a single pod by name.

  Try: “Show me the details of pod api-7d9f in staging.”

  Read · pods_get

• Reads a pod's logs

  Returns the last lines of a pod's logs, 100 by default. It can return logs from the previous container run.

  Try: “Show the last 200 log lines of the checkout pod and explain the error.”

  Read · pods_log

• Shows CPU and memory use of pods

  Reads pod CPU and memory from the Kubernetes Metrics Server.

  Try: “Which pods in prod use the most memory?”

  Read · pods_top

• Lists cluster events

  Lists warnings, errors and state changes, which helps debug why something fails.

  Try: “What warning events happened in the web namespace today?”

  Read · events_list

• Lists namespaces

  Lists every namespace in the current cluster.

  Try: “List all namespaces.”

  Read · namespaces_list

• Lists OpenShift projects

  Lists OpenShift projects in the current cluster.

  Try: “Which OpenShift projects exist on this cluster?”

  Read · projects_list

• Shows CPU and memory use of nodes

  Reads node CPU and memory from the Metrics Server for one node or all of them.

  Try: “Which node is under the most CPU load?”

  Read · nodes_top

• Shows detailed stats for one node

  Reads detailed CPU, memory, disk and network stats for a node and its pods through the kubelet.

  Try: “Give me the stats summary for node worker-2.”

  Read · nodes_stats_summary

• Reads system logs from a node

  Gets logs such as kubelet or a log file path from a node through the Kubernetes API.

  Try: “Show the last 100 kubelet log lines from node worker-2.”

  Read · nodes_log

• Lists any kind of cluster resource

  Lists resources of any type, such as Deployments, Services or Ingresses, with label and field filters.

  Try: “List all Deployments in the prod namespace.”

  Read · resources_list

• Shows one resource of any kind

  Returns a single resource by version, kind and name.

  Try: “Show me the Ingress named web in prod.”

  Read · resources_get

• Looks up the right API version for a kind

  Finds the API version, plural name and scope for a resource kind. It appears only in the current main branch, not the latest release.

  Try: “What API group is the Route kind in?”

  Read · api_resources_list

• Lists the clusters in your kubeconfig

  Lists the context names and server addresses in your kubeconfig file.

  Try: “Which clusters can you reach?”

  Files · configuration_contexts_list

• Shows your kubeconfig

  Returns your Kubernetes configuration as YAML, trimmed to the current context by default.

  Try: “Show my current kubeconfig context.”

  Files · configuration_view

• Lists available clusters or targets

  Lists the targets the server can use.

  Try: “List the available targets.”

  Read · targets_list

• Creates or updates any resource from a manifest

  Applies a full YAML or JSON manifest with server-side apply. Fields the earlier version set but the new manifest leaves out are removed.

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

  Write · resources_create_or_update

• Reads or changes the replica count

  Returns the current scale of a Deployment or StatefulSet and sets a new one if you give a number.

  Try: “Scale the web deployment in prod to 5 replicas.”

  Write · resources_scale

• Starts a pod from a container image

  Runs a pod from the image you give, and can expose a port.

  Try: “Run nginx:latest in the test namespace and expose port 80.”

  Write · pods_run

• Runs a command inside a pod

  Runs a command in a pod container and returns the output, like a shell session.

  Try: “Run "ls -l /app" inside the api pod.”

  Runs code · pods_exec

• Deletes a pod

  Deletes a pod by name. A controller usually creates a new one.

  Try: “Delete the crashing pod api-7d9f in staging.”

  Delete · pods_delete

• Deletes any resource

  Deletes a resource of any kind by name, with an optional grace period (zero deletes immediately).

  Try: “Delete the Service named old-api in staging.”

  Delete · resources_delete

• Lists Helm releases

  Lists Helm releases in a namespace or all namespaces. Needs the helm toolset turned on.

  Try: “Which Helm releases are installed in prod?”

  Read · helm_list

• Installs a Helm chart

  Installs a chart as a release in a namespace, with optional values. Needs the helm toolset turned on.

  Try: “Install the grafana chart in the monitoring namespace.”

  Write · helm_install

• Removes a Helm release

  Uninstalls a Helm release. Needs the helm toolset turned on.

  Try: “Uninstall the release test-grafana.”

  Delete · helm_uninstall

• Reads the logs of a Tekton pipeline run

  Gets logs for all task runs owned by a pipeline run. Needs the tekton toolset turned on.

  Try: “Show the logs of pipeline run build-42.”

  Read · tekton_pipelinerun_logs

• Diagnoses a failed Tekton pipeline run

  Collects read-only evidence about a failed run: conditions, failed task runs and steps, and log tails. Needs the tekton toolset turned on.

  Try: “Why did pipeline run build-42 fail?”

  Read · tekton_pipelinerun_diagnose

• Reads the logs of a Tekton task run

  Gets logs from a task run through its pod. Needs the tekton toolset turned on.

  Try: “Show the logs of task run unit-tests-7.”

  Read · tekton_taskrun_logs

• Starts a Tekton pipeline

  Starts a pipeline by creating a PipelineRun. Needs the tekton toolset turned on.

  Try: “Start the deploy pipeline.”

  Write · tekton_pipeline_start

• Starts a Tekton task

  Starts a task by creating a TaskRun. Needs the tekton toolset turned on.

  Try: “Start the lint task.”

  Write · tekton_task_start

• Restarts a Tekton task run

  Creates a new task run with the same spec. Needs the tekton toolset turned on.

  Try: “Restart task run unit-tests-7.”

  Write · tekton_taskrun_restart

• Restarts or cancels a pipeline run

  Restarts a pipeline run with the same spec or cancels it. Needs the tekton toolset turned on.

  Try: “Cancel pipeline run build-42.”

  Write · tekton_pipelinerun_lifecycle

Watch out for

• It can delete resources and run commands in pods

  The default tools include deleting any resource and running commands inside a pod. Use read-only mode, or a service account without those rights. github.com

• It can create or change any resource

  One tool applies full manifests and another scales workloads. A wrong manifest can replace live settings. github.com

• Access is your whole kubeconfig

  It can use every context in your kubeconfig. Keep production out of that file, or limit the account. github.com

• Secrets can be read unless you block them

  The README shows a denied_resources setting that blocks the Secret kind. Add it if the assistant should not see secrets. github.com

• Extra toolsets are off by default

  Helm, Tekton, Kiali, KubeVirt and the observability tools only load when you list them in the toolsets setting. github.com

Security advisories