Utilities

Runbook Generator

Generate a complete on-call runbook in Markdown. Symptoms, diagnostic commands, mitigation steps, rollback procedure, escalation path, and dashboards — all from one form. Copy or download.

Identifier this runbook is keyed off (matches your alerting rule)

Affected service name

How urgent is this

Who owns the response

Where to coordinate

What this alert means in plain language

What customers / dashboards show. One per line.

Copy-paste commands to confirm and scope the issue

Numbered steps the responder should follow

If a deploy is implicated

Who to escalate to and when

One URL per line

Other alerts that often fire alongside this one

Fill in the fields above and click Generate Runbook.

The generated Markdown will appear here, ready to copy or download as a .md file.

Automate your incident response

Reduce MTTR by 90% with AI-powered root cause analysis. Free to start.

Try Uptimes.ai Free

The 3-AM rule of runbook writing

Imagine you have just been paged at 3 AM. You do not own this service. You have never seen this alert before. You have five minutes to act before the on-call rotation gets escalated. What does the runbook need to tell you?

That mental model produces the right structure: a one-line summary (what does this mean?), a few diagnostic commands (is this real and what is the scope?), numbered mitigation steps (what do I do?), a rollback procedure (if a deploy is implicated), an escalation path (when to wake somebody else up), and links to dashboards. Anything else is footnotes — keep the main path linear and copy-pasteable.

What separates great runbooks from mediocre ones

  • Diagnostic commands you can paste — not "check the dashboard" but the actual kubectl, curl, or PromQL query. The responder should be able to confirm the issue in under a minute.
  • Numbered mitigation steps in order. Branching is OK ("if X then Y else Z") but always commit to a default path.
  • A rollback procedure with the exact command. Most production incidents are deploy-related; rollback being one paste away pays for itself.
  • Explicit escalation triggers — not "escalate if it gets bad," but "page Tier 2 after 30 minutes if not resolved." Time-bound and unambiguous.

Runbooks and AI SRE agents

Runbooks used to be just for humans. Now AI SRE agents read them too — Uptimes.ai\'s agent ingests linked runbooks during investigation and uses them as candidate actions. This means a well-structured runbook now has compounding value: humans benefit, the agent benefits, and the agent\'s automated remediations can be derived directly from your mitigation section.

This is a meaningful shift in how to think about runbook quality. The same clarity that helps a sleepy human at 3 AM also helps an automated system reason about what to do — and bad runbooks become a bottleneck for both.

Frequently Asked Questions

What is a good runbook?+
A good runbook tells the responder, "this alert means X, here is how to confirm, here is how to fix it." It assumes the reader has been paged at 3 AM, has 5 minutes of context, and needs to act fast. The best runbooks have copy-pasteable diagnostic commands, numbered mitigation steps, an explicit rollback procedure, and a clear escalation path. The worst runbooks are essay-length pages of context with no actionable steps.
Should every alert have a runbook?+
Every alert that pages a human, yes. If you cannot write a runbook for an alert — because the response is "look at the dashboard and figure it out" — that is usually a sign the alert is not actionable and should be deleted, downgraded to a ticket, or refined until you can describe what to do.
How do I keep runbooks from going stale?+
Three habits: (1) link the runbook from the alert annotation so responders see it during incidents and notice when it is wrong, (2) add a "last updated" date and treat anything over 6 months old as suspect, (3) update the runbook as part of every postmortem that involved it. The Uptimes.ai postmortem template includes a runbook-update action item by default.
Where should I store runbooks?+
Wherever your team already lives — Confluence, Notion, your wiki, or a /runbooks directory in the repo. The most important property is that the URL is permanent and linkable from your alerting system. Repo-based runbooks have the advantage of being version-controlled alongside the code that produces the alerts.
Does Uptimes.ai use these runbooks?+
Yes — Uptimes.ai's AI SRE agent reads runbooks from your service catalog (via the Backstage integration) and uses them as input when investigating an incident. If your alert has a documented mitigation, the agent considers it as a candidate action. Well-written runbooks make the agent dramatically more effective.