Page the right person. Let everyone else sleep.

Relay routes every alert from your monitoring to the one engineer who owns it, using rules you write in plain language and test before they go live.

Start free — up to 5 responders Walk through a real 2 a.m. incident

Imports your PagerDuty or Opsgenie schedules in about two minutes.

Used by teams at

  • tallyline
  • FERNHILL
  • quayside/
  • Brightpost
  • Lumen Bay
  • Corvid

Eight engineers, forty services, one phone at 2 a.m.

Relay is built for teams of 3 to 50 without a dedicated ops rotation: every rule reads like a sentence, and you can test it against a real alert before it wakes anyone up.

Routing rules you can read aloud

Rules run top to bottom; the first match decides who gets called, how, and who's next if nobody answers.

payments.rules 5 rules · edited by Mara Okafor, Oct 9
  1. 01 If severity is SEV1 on checkout-api, call the payments on-call; escalate to secondary after 5m.
  2. 02 If severity is SEV1 on any other service, call the platform on-call.
  3. 03 If severity is SEV2 during Mon–Fri 09:00–18:00, post to #incidents and push the service owner.
  4. 04 If severity is SEV2 outside those hours, push the team's on-call; call only if still open after 15m.
  5. 05 Everything else opens a Linear ticket for the owning team. Nobody is paged.
SimulatorTest an alert against these rules
Service
Severity
Time
Calls Dev Patel by phone
Payments on-call. Escalates to Mara Okafor at 02:21 if unacknowledged.
matched rule 1

Schedules with overrides that stick

Weekly rotations and one-off swaps. An override never edits the base schedule, so next week is still right.

Primary Secondary Override · Sat swap

Now, Tue 02:14: Dev Patel primary, Mara Okafor secondary.

Fourteen alerts, one phone call

Relay folds alerts by service and time window, so a cascading failure rings once.

14 → 1alerts to incidents on INC-2041
−91%pages vs. last quarter
3m 40smedian time to acknowledge
0missed pages in 90 days
  • Datadog
  • Grafana
  • Prometheus
  • Sentry
  • CloudWatch
  • GitHub
  • Slack
  • Linear

Every signal for the incident, on one screen.

INC-2041 as Dev saw it at 02:20, six minutes after the first alert and one minute after he picked up.

SEV1 INC-2041 Acknowledged
Checkout API p95 latency above 2s
Roll back #1187Resolve
2
p95 latency · checkout-api · last 30mthreshold 2.0snow 2.48s
deploy #1187 alert 02:14 2.0s
01:5002:0002:1002:20
1
14 alerts grouped · Datadog and Sentry · 02:14–02:20
  • p95 latency > 2s×6
  • HTTP 5xx > 2%×4
  • payment-queue depth×3
  • TaxClientTimeout×1
  1. Deploy #1187 shipped · v2.41.0 by Lena Ruiz
  2. p95 latency hits 2.4s · Datadog
  3. Called Dev Patel · rule 1
  4. Dev acknowledged · escalation to Mara cancelled
  5. Sam Reyes joined from Slack

  1. Grouped alerts

    Fourteen alerts from Datadog and Sentry folded into one incident, so Dev got one call instead of fourteen.

  2. Latency against your threshold

    p95 plotted against the 2s line you set, with the deploy marked where the climb begins.

  3. Who's on it

    Who was called, who acknowledged and when, and who joined from Slack. Mara's escalation cancelled itself.

  4. The deploy that did it

    Relay matches the incident to the last deploy on checkout-api and puts the rollback command next to it.

Teams that stopped dreading their phone.

Small teams moved to Relay from PagerDuty, Opsgenie and a shared Slack channel nobody muted.

“We're eight engineers covering forty services. Whoever held the pager got around 140 pages a month, and most of them fixed themselves. We wrote eleven rules in an afternoon. Now it's twelve pages a month, and every one is real.”
140 → 12pages per month
14m → 3mmedian time to acknowledge
PN Priya NatarajanHead of Engineering, Tallyline
“The simulator sold us. We replayed last quarter's alerts against every rule before we switched off PagerDuty.”
TH Tomás HerreraStaff SRE, Quayside
“When my co-founder went on parental leave, I covered her week with two overrides. The base schedule never changed.”
AB Aiko BrandtCTO, Fernhill

Pay for people on call. Nobody else.

A seat is anyone who can be paged. Stakeholders, managers and Slack watchers who follow incidents are free.

Free

For side projects and new teams

$0forever

Up to 5 responders, no card needed

Start free
  • 5 routing rules
  • Push and email alerts
  • 1 schedule with overrides
  • 3 integrations

Team

Recommended

For teams with a real rotation

$12/ responder / month

Billed yearly, or $15 month to month

Start a 14-day trial
  • Unlimited rules and the simulator
  • Phone and SMS with escalation
  • Alert grouping by service
  • Deploy correlation from GitHub
  • Slack incident channels

Business

For several teams sharing on-call

$24/ responder / month

Billed yearly, invoicing available

Talk to us
  • Everything in Team
  • SAML SSO and SCIM
  • Cross-team escalation policies
  • Audit log kept for 1 year
  • 1-hour response on SEV1 support

Compare plans

FeatureFreeTeamBusiness
RespondersUp to 5UnlimitedUnlimited
Routing rules5UnlimitedUnlimited
Integrations3UnlimitedUnlimited
Phone and SMS—Not includedIncludedIncluded
Alert groupingBy serviceService and time windowService and time window
Deploy correlation—Not includedIncludedIncluded
Slack incident channels—Not includedIncludedIncluded
SAML SSO and SCIM—Not included—Not includedIncluded
Audit log—Not included30 days1 year
SupportCommunityEmail, next business day1-hour SEV1 response

Route your next 2 a.m. page right.

Connect your monitoring, import your schedule, write your first rule. Most teams are done in ten minutes.