Kubernaut — Kubernetes desktop app

Why

Kubernetes GUIs tend to break one of two ways: they hold the entire object list in the frontend and choke on a large cluster, or they read ~/.kube/config on launch and put every configured cluster one click away, production included. Kubernaut keeps all cluster state in Rust — the UI only ever asks for a page or subscribes to a delta stream — and starts with no clusters connected; every one is added explicitly.

Screenshots

Taken against a local k3d cluster, not a real one — every name in these images is invented.

Cluster overview: gauges, usage history, live issues — Cluster overview, with an issue an `ImagePullBackOff` pod put there itself

Deployments table with row selection — Row selection on any resource table

Diagnose tab quoting the kubelet's own error — Diagnose: the cluster's own error, and what to do about it

Bulk action bar: delete, restart, export — Act on a selection — delete, restart, export

Create dialog, form mode, for a Deployment — Create, as a form or as YAML

Ingress form with backend service and port chosen from the cluster — Reference fields read from the cluster, not typed

Helm releases and revision history — Every Helm release, read straight from the cluster

Security Center workload findings — Workload posture findings, no scanner required

Container shell terminal — A shell into the pod

Features

Multi-cluster

Contexts added explicitly, lazy health-probed connections, protected contexts that refuse destructive actions outright.

Resource browser

Full API discovery including CRDs, virtualised tables, row selection, a fuzzy command palette across clusters.

Create, edit, import

Form or YAML with a real cluster-side diff before anything applies; reference fields read from the cluster, not typed.

Diagnose

Reads a pod's status the way a human would — quotes the cluster, not a guess — with a next step straight into logs or a shell.

Logs and terminal

Multi-pod log tail, four terminal modes, port forwarding bound to loopback by default.

Metrics and topology

Cluster gauges, namespace heatmap, sizing recommendations, and an Ingress → Service → Pod → Node graph.

Helm

Every release read straight from the cluster's own Secrets, with a real pre-upgrade diff.

Security Center

Workload posture checks, RBAC analysis, and image CVEs — no scanner required.