Monitoring and alerts
Know when a server needs you, with nothing to install.
Every minute K2Shell signs in with a short-lived certificate, runs read-only commands, and charts what it finds. Alerts go to email, Telegram or a webhook.
What you get
Charts
CPU, memory, swap, load, disks, inodes, network and disk I/O over 1 hour to 30 days.
Health checks
Failed systemd services, restarting containers, OOM kills and clock drift.
TLS expiry
Certificates on the server checked and alerted on before they lapse.
Alerts
Thresholds you set, alerts that clear themselves, batching so you are not flooded.
Mutes
Silence a server or an alert for a while during maintenance.
90 days kept
History is kept for 90 days and available to Claude through metrics_get.
Questions
Do I install an agent?
No. Monitoring uses the same certificate login as everything else, with read-only commands.
Which systems are supported?
Linux servers with a POSIX shell. Most metrics come from /proc, df and systemctl.