Need more than 2? $39 once unlocks unlimited servers. Yours to keep, no account.
CPU, RAM, disk for every mounted filesystem, load average, uptime. Updated anywhere from every second to every five minutes. Adaptive polling speeds up when a metric warms up or recovers, then falls back to your configured interval.
Will it work on my distro? If the box has ssh, /proc, and df, Kitaso reads it. Overlay, tmpfs, devtmpfs and /boot/efi are filtered out automatically.
Want history while you sleep? A small bash + cron + sqlite collector installs in one click over the SSH connection Kitaso already has. It runs 24/7. Kitaso auto-imports its data on launch and every five minutes after, deduplicated against live polling so charts stay clean.
What does it cost on the server? About 50 KB on disk to start, a cron job every minute, and an idle bash process. No persistent listener. No open ports.
Per-container CPU and memory tracking. Running and stopped state, image tag, and the status line Docker reports. Expandable per-container charts. Image tag changes are recorded as deployment events and marked on the chart, so a CPU spike has a story attached.
Do I need to expose docker.sock? No. Kitaso runs docker stats over the same SSH session it's already using. The socket stays where it belongs.
Warning and critical thresholds per metric, per server. Warm-up and recovery windows so spikes have to sustain before firing and clear before resolving. Severity escalation (warning → critical) fires immediately without a fresh warm-up. Per-server snooze from 5 minutes to "until tomorrow."
Won't it be spammy? No. Disk alerts deduplicate by underlying device, so a dozen Docker overlay mounts on one disk produce one alert, not twelve.
One SQLite database in WAL mode, in your Application Support folder. Configurable retention from 1 to 365 days, with on-demand Prune Now. No account, no cloud sync. Configuration files are written with 0600 permissions.
Export when you need it. Metric history to CSV, configuration to JSON. Yours to take anywhere.
Time ranges from 1 hour to 30 days. Drag, scroll-wheel, or arrow-key navigation with snap-to-now. All charts on the page pan together. LTTB downsampling so weeks of data stay responsive. Hover crosshair tooltip with exact date, time, values, percentages.
No fake trends. Genuine gap detection. If collection actually stopped, the line breaks and the tooltip says "no data at cursor" instead of interpolating something that didn't happen.
No subscription · No cloud
Questions? [email protected]