v0.1.0-beta.26

Released: March 19, 2026

What's New

Resource Actions

You can now perform common Kubernetes operations directly from Kunobi without switching to the terminal:

Rollout Restart (Shift+R) — zero-downtime rolling restart for Deployments, StatefulSets, and DaemonSets
Suspend / Resume (p) — toggle suspension on CronJobs and Flux resources (HelmRelease, Kustomization, GitRepository)
Trigger Job (j) — create a one-off Job from any CronJob's template
Drain Node (d) — cordon and evict all pods with live progress tracking, respecting PodDisruptionBudgets
Rollback Deployment (Shift+B) — browse ReplicaSet revision history with container images and change-cause annotations, then rollback to any previous revision

All actions use confirmation dialogs with Enter key support and automatic focus restoration to the table row after completion.

Log File Viewer

A new built-in log viewer powered by a virtual TUI engine lets you browse log files directly inside Kunobi:

Full-text search (⌘F / Ctrl+F) and regex filtering (/) with match navigation (n / Shift+N)
Word wrap toggle (w) and horizontal scrolling (Shift+Wheel)
Vim-style navigation (j/k to scroll, g/G for top/bottom)
Follow mode (f) to auto-scroll to new log lines as they arrive
Mouse wheel scrolling with a draggable scrollbar
ANSI escape codes are automatically stripped for clean output
All shortcuts registered in the keyboard shortcuts help panel

Pod Metrics History

Pod metrics now include historical data with interactive sparkline charts, backed by automatic PromQL discovery:

Five chart panels — CPU usage, memory usage, filesystem usage, filesystem I/O, and network throughput
Automatic backend discovery — finds Prometheus, VictoriaMetrics, Thanos, and SigNoz instances in your cluster
VictoriaMetrics cluster mode — correctly discovers vmselect services in multi-tenant cluster deployments
Sparkline previews — inline mini-charts in the pod table columns show trends at a glance

Recent Clusters

Kunobi now tracks your last five connected clusters and makes reconnecting instant:

Quick-connect banner — shows recent clusters on the Clusters page and the K8s drilldown overlay, with one-click connect/disconnect
Smarter sorting — the cluster selector now orders: favorites first, then connected, then recent (by last connected time), then the rest

Git History

Browse your cluster's GitOps repository history without leaving Kunobi:

Clone and browse — Kunobi clones your GitOps repo when you open the Git History tab and displays commit history with ref badges
Structured error handling — clear error cards with actionable hints when authentication fails or tools are missing, including a one-click terminal login flow
Fetch warning banner — shows when the last fetch encountered issues, with a refresh button
Re-clone support — re-clone from scratch if the local copy gets into a bad state

Unified Filter Bar

All pages — Clusters, Settings, and K8s drilldown — now use the same FilterBar component for search and filtering. This replaces the old search bar with a consistent experience across the app, including smart filter pills, keyboard navigation (Arrow Down to focus results), and per-tab scoped state.

MCP Universal Install

The MCP settings page now offers a single npx @kunobi/mcp install command that auto-configures all detected AI clients (Claude, Codex, Gemini) at once, replacing the previous per-client setup. Includes a run-in-terminal button and migration notice for existing installations.

Improvements

K8s table performance — store-backed row rendering eliminates unnecessary React re-renders when resource data updates, resulting in smoother scrolling and faster table updates on large clusters.

Bug Fixes

Tab state isolation — sidebar, resource filter, and search bar state are now properly scoped per tab. Switching tabs no longer leaks state between views.
Sidebar closes on disconnect — the left sidebar now correctly closes when a cluster disconnects from another tab, instead of remaining open with stale data.
Namespace drilldown fixed — the FilterBar no longer creates spurious namespace pills during drilldown navigation, which was filtering out cluster-scoped resources and resetting the view stack.
Cluster disconnect stuck — orphaned ping loops from duplicate connection attempts no longer override the disconnect status, fixing clusters that appeared impossible to disconnect.
JWT session expiry — long-running sessions now detect expired tokens before API calls and trigger immediate refresh, with telemetry exporters rebuilt using fresh credentials.
Blurry terminal on Linux — terminal text on X11 with custom DPI settings is now rendered at the correct resolution.
Cluster card overflow — long cluster IDs and error messages are now truncated with tooltips instead of breaking the card layout.
MCP session stability — hardened session transport handling prevents stale sessions from accumulating.
Windows command windows flashing — background processes (tool detection, Docker checks, Git operations, Mise installs) no longer spawn visible console windows on Windows during app startup and navigation.

Installation

Feedback

We'd love to hear your thoughts!