Summary
Cluster triage in conversation, with one call that writes.
Listing, describing and reading logs cover the debugging loop and change nothing, which is what makes this attachable to a live cluster. The exception is the rolling restart, which does act on a deployment — worth knowing before you point it at production.
What it is
A Go MCP server for Kubernetes inspection. It reads through your existing kubeconfig, works with any Custom Resource Definition in the cluster, and finds resources by API-group substring — flux for FluxCD, argo for ArgoCD.
What you get
- Resources of any kind listed and filtered, custom resources included
- A named resource described in detail
- Pod logs retrieved, with filtering
- Discovery by API-group substring, so the exact group name is not needed
- A rolling restart of a deployment
Requirements
A kubeconfig with read permission on whatever you want to inspect — it picks up ~/.kube/config by default, or the path in KUBECONFIG. Install with go install, or build from source with Go 1.24 or later.
Setup effort
One command — go install github.com/k4mrul/kubernetes-mcp@latest