Kubeview

KubeView MCP Server provides a read-only Model Context Protocol interface that lets AI agents safely inspect and diagnose Kubernetes clusters. It exposes Kubernetes, Helm, Argo Workflows, and Argo CD data in a structured, read-only manner to help you debug and diagnose issues without risking accidental changes.

How to use

You interact with KubeView MCP through an MCP client such as Cursor IDE or Claude Code CLI. Start the MCP server locally and point your client to the server using the commands shown below. You can run this in combination with clients to remotely inspect resources, fetch metrics, view logs, and explore Helm releases and Argo resources without making any changes to your cluster.

How to install

Prerequisites you need before starting include Node.js version 18 or newer and access to a Kubernetes cluster. It is also helpful to have Helm, Argo, and Argo CD CLIs available in your PATH.

# Prerequisites check
node -v
kubectl version --client
helm version
argocd version

# Install and start the MCP server via npx
npx -y kubeview-mcp

# Optional: add Kubernetes MCP server to Claude or Cursor workflows
claude mcp add kubernetes -- npx kubeview-mcp

Additional setup and configuration

Configure your MCP client to use KubeView by adding a server entry that points to the local MCP instance. You can specify the command to run the MCP server through your client’s MCP configuration.

{
  "mcpServers": {
    "kubeview": {
      "command": "npx",
      "args": ["-y", "kubeview-mcp"]
    }
  }
}

Environment and runtime configuration

The server can be configured using the following environment variables. These control how the MCP server operates and how much data is masked in outputs.

| Variable | Description | Default |
|----------|-------------|---------|
| KUBECONFIG | Path to kubeconfig file | ~/.kube/config |
| MCP_MODE | Server mode: all, code, or tools | all |
| MCP_LOG_LEVEL | Log level: error, warn, info, debug | info |
| MCP_HIDE_SENSITIVE | Enable global sensitive data masking | false |

Code mode and secure execution

Code Mode provides a sandboxed TypeScript environment to perform complex reasoning and multi-step workflows. It includes a dynamic TypeScript API surface, tool search utilities, and a guarded VM sandbox to ensure safe execution while you reason about cluster state.

To enable Code Mode, set the environment variable MCP_MODE to code and use the code-mode prompts in your MCP client.

Available tools

kube_list

List Kubernetes resources and provide cluster diagnostics. Use to enumerate namespaces, pods, services, and other resources.

kube_get

Describe specific Kubernetes resources by type and name to understand status, spec, and metadata.

kube_metrics

Fetch CPU and memory metrics for nodes and pods to aid capacity planning and performance troubleshooting.

kube_logs

Fetch or stream container logs for debugging and incident investigation.

kube_exec

Execute read-only commands in containers to inspect state without modifying the cluster.

kube_port

Port-forward to Kubernetes services or pods for local debugging and access testing.

kube_net

In-cluster network diagnostics to identify connectivity issues.

helm_list

List Helm releases deployed in the cluster.

helm_get

Fetch release values, manifests, and history for Helm releases.

argo_list

Manage and inspect Argo Workflows through the Kubernetes API.

argo_get

Retrieve details for specific Argo Workflows.

argocd_app

Inspect Argo CD applications and their Kubernetes resources.

run_code

Execute sandboxed TypeScript code to perform complex tasks and data processing.

plan_step

Record and manage step-by-step planning state to support long investigations without bloating chat context.