We stopped waking up the whole team.
Before Relay every alert went to a channel and whoever was awake took it. Now one person is paged, with the runbook. Our on-call survey went from dread to fine.
Features
Everything the 3am screen needs.
Four parts, one screen. Jump to one, or read down.
Jump to a feature
The four parts
Alerts that find a person
One page to the engineer who owns the service, escalating every five minutes until someone answers.
Runbooks that open themselves
The fix for this alert, on the screen that shows the alert, with last time's notes.
A status page customers trust
Updated from the incident, in your words, before the first support ticket.
Postmortems in a day, not a month
The timeline writes itself from the chat, the pages and the deploys.
Alerts that find a person
One page to the engineer who owns the service, escalating every five minutes until someone answers.
Relay reads your service map and pages the owner, not a rota. If they do not acknowledge in five minutes it moves up the chain, then to everyone. Pages arrive by push, call and text; the call reads the alert aloud.
Runbooks that open themselves
The fix for this alert, on the screen that shows the alert, with last time's notes.
Every alert carries its runbook: the steps, the dashboards, the people to call, and what the last three engineers did when it fired. Runbooks are Markdown in your repository, so they are reviewed like code.
A status page customers trust
Updated from the incident, in your words, before the first support ticket.
Your status page is written from inside the incident, in a template your team agreed on in advance. Subscribers get the first update within ninety seconds and the all-clear with the cause, not a shrug.
Postmortems in a day, not a month
The timeline writes itself from the chat, the pages and the deploys.
Relay keeps the timeline as the incident happens: every page, every message, every deploy and rollback, with the clock. The postmortem starts with the facts filled in, so the meeting is about what to change.
How it works
Four steps, then sleep.
Connect your alerts
Datadog, Grafana, Sentry, CloudWatch or a webhook. Ten minutes, no agent.
Map your services
Who owns what, in a file in your repo. Relay reads it and keeps it current.
Write the first runbook
Start from ours. Relay opens it the next time the alert fires.
Sleep
The right person is paged, the status page speaks, and the timeline is kept. You read it in the morning.
Beta teams
What they say after a quarter.
Priya Natarajan, CTO, Tally, Beta customer since May 2026 The status page is finally honest.
We used to update it after the fact. Now it is written from inside the incident and customers see it before they notice anything. Support tickets during incidents fell by half.
Marcus Oyelaran, head of platform, Ferry, Beta customer since June 2026 Postmortems take an hour.
The timeline is already there, with the clock. We argue about what to change instead of what happened.
Lena Hartmann, engineering manager, Grove, Beta customer since July 2026