Release notes

First release

• Sign in with a free KubeCrow account, with Google or GitHub: click Sign in, pick one in your browser, and KubeCrow opens on your clusters. The same email on both is the same account. My account… (top right) opens your account page, and KubeCrow keeps working offline for 7 days. Your account keeps only your name, email, photo and plan; your clusters and their data never leave your Mac.
• Fast on big, busy clusters: after the first full list, KubeCrow sends the window only what changed.
• A menu bar icon with the number of critical problems, and a menu of your connected clusters, worst first. Closing the window keeps KubeCrow running there, so clusters stay connected and alerts keep coming. Open at login starts it quietly in the menu bar.
• Every section: Cluster, Workloads, Config, Network, Storage, Access Control, Helm and Custom Resources, with live updates, sorting, search and CSV export.
• AWS EKS discovery across all your AWS profiles and enabled regions, grouped by account.
• Production guard: mark clusters Protected or Read-only; destructive actions ask you to type the cluster name.
• Health page with Problems, cluster CPU and memory history (from Prometheus, or sampled by KubeCrow and kept for 24 hours) and background alerts with quiet hours.
• Stuck rollouts flagged in Problems and alerts, even while the old pods keep the workload fully ready, and rolled back in one click from Problems or the pod’s Diagnose tab, which also names the new pod that’s failing and why.
• Right-sizing: each container’s CPU and memory requests and limits against 24 hours or 7 days of real use from Prometheus, with suggested values, the containers close to being OOM-killed, and a ready-to-copy kubectl set resources command.
• What changed: a timeline of rollouts, rollbacks, scaling, restarts, Job results and warnings, plus ConfigMap edits with their old and new values and Secret edits (key names only, never values).
• Search logs: one search across every pod in your namespaces through the cluster’s Loki, including pods that were replaced or deleted, also one click from a crashing pod’s error. Multi-tenant Loki and Mimir work without setup: KubeCrow finds the tenant in your Alloy, Promtail, Fluent Bit, Vector, OpenTelemetry or Prometheus config.
• Logs with time ranges, timestamps, search and pretty-printed JSON.
• Pod and workload metrics from Prometheus over 1 hour to 7 days, against requests and limits.
• Pod and node shells, a Debug shell for images that have no shell of their own, and port forwarding that offers to restore last time’s forwards after a restart, on the same local ports.
• Compare clusters: every workload’s image and replicas side by side, how far apart builds are (“7 builds behind”), the kubectl set image command to bring one cluster to the other’s version, and, where a workload isn’t deployed, that cluster’s version of the same image.
• YAML editor for any resource; inline ConfigMap and Secret value editing.
• Helm charts and releases via your helm CLI: install, upgrade and roll back to any revision.
• Permission-aware actions: anything you aren’t allowed to do, including bulk actions, is greyed out with the reason.
• Running images per workload, favourites, faster namespace switching, and a Settings menu with per-cluster names, colours and start namespace, plus settings export and import.
• Ocean, Midnight and Classic themes following macOS light and dark, and a ⌘K command palette.
• Check for updates (Settings → About) tells you when a new version is out, with a Download button, and KubeCrow shows what’s new once after an update. Help → Report a Problem… opens a GitHub issue with your version filled in.
• Universal build for Apple Silicon and Intel Macs. Not yet notarised; see install notes.