MCP Server that can connect to a Kubernetes cluster and manage it. Supports loading kubeconfig from multiple sources in priority order.
MCP.Pizza Chef: Flux159
Your assistant talks to whatever cluster kubectl already points at, so it can list resources, describe them, read logs, scale a deployment, apply a manifest, restart a rollout or clear out stuck pods. A guided troubleshooting routine walks through diagnosing a failing workload step by step, and Helm charts can be installed or upgraded the same way. There is nothing extra to sign up for — it reuses the cluster access you already have on your machine.
Coding
Use This MCP server To
Ask which pods are failing and why Read the logs from a crashing container Scale a deployment up before a traffic spike Restart a rollout that got stuck Clean up pods left in a broken state Install or upgrade a Helm chart by asking Walk through diagnosing a failing workload step by step
README
mcp-server-kubernetes FAQ
What do I need before it works?
kubectl installed, a working kubeconfig, and access to a cluster. Helm is optional and only needed for chart commands.
Which apps does it work in?
Claude Desktop, Claude Code, Codex, Cursor and VS Code, and Claude Desktop also lists it as an installable extension.
Do I need an account key?
No — it reuses the cluster credentials already on your machine, so there is nothing extra to sign up for.
Can I use this to find out why a pod keeps restarting?
Yes — it can describe the pod, read its logs and follow a built-in troubleshooting routine to narrow down the cause.
Can it change things, or only look?
It can create, apply, delete, scale and patch resources, so point it at a cluster you are comfortable letting it touch.
How hard is setup?
One command in Claude Code or Codex, or a short config block elsewhere — but you should already be comfortable with kubectl.