Model Context Protocol moderate risk

Kubernetes

This MCP server enables interaction with Kubernetes clusters, offering both read and write operations, including resource management, diagnostics, and Helm chart deployment.

Connections & Capabilities

Connects To

GitHubDockerKubernetes

Capabilities

readwriteexecadmin

Quickstart

Install

npx mcp-chat --server "npx mcp-server-kubernetes"

Config

{
  "mcpServers": {
    "kubernetes": {
      "command": "npx",
      "args": ["mcp-server-kubernetes"]
    }
  }
}

Exposed MCP Tools (15)

safe

kubectl_get

Retrieves Kubernetes resources.

Read-only operation.

safe

kubectl_describe

Describes Kubernetes resources.

Read-only operation.

moderate

kubectl_create

Creates Kubernetes resources.

Creates new resources, but doesn't modify existing ones without further action.

moderate

kubectl_apply

Applies YAML manifests to create or update resources.

Can modify existing resources based on the manifest.

critical

kubectl_delete

Deletes Kubernetes resources.

Destructive operation that permanently removes resources.

safe

kubectl_logs

Retrieves logs from Kubernetes pods.

Read-only operation.

moderate

kubectl_scale

Scales the number of replicas in a deployment or replicaset.

Modifies the number of replicas, potentially impacting application availability.

high

kubectl_patch

Updates specific fields of a Kubernetes resource.

Can modify resource configurations, potentially causing unexpected behavior.

moderate

kubectl_rollout

Manages deployment rollouts (status, history, pause, resume, undo).

Can alter the state of deployments, potentially disrupting service.

critical

kubectl_generic

Executes arbitrary kubectl commands.

Unrestricted access to kubectl functionality, including destructive operations.

moderate

helm_template_apply

Applies a Helm chart template to the cluster.

Can create or modify resources based on the template.

high

helm_template_uninstall

Uninstalls a Helm chart template from the cluster.

Removes resources deployed by the Helm chart.

high

cleanup_pods

Cleans up problematic pods in various error states.

Deletes pods, potentially disrupting workloads.

high

node_management

Cordons, drains, and uncordons Kubernetes nodes.

Can disrupt workloads running on the affected nodes.

safe

k8s-diagnose

Guides through a systematic Kubernetes troubleshooting flow for pods.

Primarily read-only, guiding the user through diagnostic steps.

Safety Assessment

This server provides powerful Kubernetes management capabilities, but its safety depends heavily on configuration and usage. Running in non-destructive mode significantly reduces risk. Exercise caution with `kubectl_generic` and untrusted Helm charts.

Non-destructive mode disables risky operations.
Secrets masking protects sensitive data.
RBAC in Kubernetes limits access based on configured permissions.
Troubleshooting prompt guides users through safe diagnostic steps.

Default mode allows destructive operations like deleting resources.
Improperly configured kubeconfig can grant excessive permissions.
Helm operations can introduce vulnerabilities if charts are untrusted.
kubectl_generic tool allows execution of arbitrary kubectl commands.