v1.4.0

Released: June 25, 2026

v1.4.0 promotes the 1.4 line to stable, combining everything from the v1.4.0-rc.1 release candidate after a round of soak time. The headline is a human-in-the-loop firewall that puts you in control of everything an AI assistant does to your clusters, backed by a full audit log. This release also brings per-container pod logs, a new ArgoCD Status tab and Refresh actions, freely renamable tabs, and a batch of graph-canvas and live-data reliability fixes. If you've been on the unstable channel since the release candidate, this update is effectively a no-op. New users on stable land here directly.

AI Cluster Firewall

A human-in-the-loop firewall that puts you in control of everything an AI assistant does to your clusters, backed by a complete audit trail.

AI changes your clusters only with your approval — AI assistants can scale, restart, cordon/drain, edit, delete, and reconcile resources through the kctl tool, but every mutation at or above your configured danger threshold now raises an approval popup (Approve · Reject · Approve for X minutes). It's off by default — turn it on in Settings → AI.
AI cluster connections are now visible and gated — the first time an assistant touches a disconnected cluster it raises a "Connect to cluster" approval; once approved, that cluster appears in the selector exactly like a manual connection instead of opening a hidden one.
New /audit page — every AI tool execution is recorded with cluster, resource, tool, status (success / failed / denied / dry-run), danger level and duration. Filter by status or danger, then export or clear the log.
Clear denials — when an AI tool is blocked, the response carries a structured reason and, where relevant, a direct upgrade link that matches the environment you're on, so connected assistants can explain the right next step instead of guessing.

What's New

Per-container pod logs — the pod log viewer gains a selector to switch between log streams for individual containers, including init containers that were previously inaccessible — much easier to debug multi-container pods and startup issues.
ArgoCD Status tab — Application resources gain a Status tab that surfaces the current and last sync operation plus per-resource sync results, mirroring ArgoCD's native UI alongside the existing Overview and Git History tabs.
Refresh ArgoCD Applications and ApplicationSets — a new Refresh action (with an optional hard refresh) now works directly from the Kubernetes drilldown, closing the parity gap where ApplicationSets only supported tab navigation. Bound to the f shortcut.
Rename any tab — double-click a tab label, press F2, or use the context-menu Rename to give a tab any name you like. Custom names survive navigation, and a one-click Reset name returns to the auto-generated title.

Bug Fixes

Smoother graph canvas — the graph no longer tears down and rebuilds its GPU layer on every surrounding render, and a batch of GPU/memory leaks and interaction glitches were fixed, for steadier panning, zooming and node interaction.
More reliable live cluster data — several data-sync fixes keep your live views consistent through reconnects and resyncs: live updates are now buffered until the first full snapshot arrives, stale out-of-order updates after a resync are rejected, the initial subscription can't stall on a connection blip, and views recover correctly even while hidden in a background tab.
Correct graph layouts — five verified layout bugs in the relationship graph were fixed for more accurate node and edge placement.
Watcher startup can't hang silently — Kubernetes watcher initialization is now bounded by a logged watchdog, so a stuck watch surfaces instead of leaving a view waiting indefinitely.
Service selector shows again — the Service overview now renders the selector that shows which pods a Service targets, instead of an empty section.

Installation

Feedback

We'd love to hear your thoughts!