Stasys
Monitor every server. Page the right person. Prove it was handled. One agent for Linux and Windows, with monitoring, on-call, security and compliance in one place, and a status page that stays up when yours do not.
Escalation, schedules and silences
How an alert reaches the right person, and how to stop it waking anyone for no reason.
Monitoring that tells you something is wrong and stops there has done the easy half. The gap between an alert and somebody acting on it is where outages live.
Escalation chains
A chain is chosen by severity and label, so a disk filling on a test server and a mail server refusing connections do not wake the same person at the same hour.
Each step pages the people, or the on-call schedule, it names: by phone call, text, email or push. Calls and texts go through your own Bird, Telnyx or Twilio account (Settings → Telephony); until one is set up, those pages go by email and push.
If nobody acknowledges in time, it moves to the next step, and you choose whether the chain starts again once it reaches the end. If you use the AI staff, an AI engineer can acknowledge an incident it takes on, which holds the chain like any acknowledgement. You choose this under AI staff → Action autonomy.
On-call schedules
Who is on, and when. Chains follow the schedule, so changing the rota never means rewriting the chains.
Holiday mode. When someone is away, name a backup on the On-call rotations page. While holiday mode is on, every page that reaches the person who is away also reaches the backup. The backup is added, never swapped in.
Silences and maintenance
For work you already know about. A silence stops the paging. Alerts sent in from your other tools during it are still kept on the incident history; Stasys's own threshold rules are not checked for a silenced server.
During a maintenance window, alerts from sources you set to respect maintenance do not page anyone. A critical alert always pages.
Less noise
- Mark test and staging servers as such (server page → Settings → Environment). Their incidents are raised at warning at most, except known-exploited vulnerabilities, malware and confirmed intrusions.
- One problem is one incident. A condition that keeps coming back adds to its open incident, or re-opens it within 6 hours, and pages again only if it got worse.
Heartbeats
The failure that never alerts is the one where the thing that would have alerted is the thing that died. A heartbeat expects to hear from something on a schedule and fires when it does not: the scheduled job that stopped running, or the machine that is now off. When a large part of your fleet goes quiet at once, you get one alert, not one per machine.
Last updated 3 November 2026