01 / INGEST
Everything lands in one stream
Point Datadog, Sentry, CloudWatch, Grafana and your own webhooks at Lunar. Setup is a URL and a token, not a migration.
Keeping the lights on for
The 3am problem
We looked at 4.1 million alerts across 300 engineering teams. Nine out of ten pages that wake someone up are noise, a duplicate, or something that fixes itself before anyone finishes typing their password.
How it works
Lunar sits between your monitoring and your phone. Nothing else changes.
01 / INGEST
Point Datadog, Sentry, CloudWatch, Grafana and your own webhooks at Lunar. Setup is a URL and a token, not a migration.
02 / TRIAGE
It collapses duplicates, correlates against the last deploy, checks dependent services, and matches the signature to a runbook you already wrote.
03 / RESOLVE OR WAKE
When a runbook applies, Lunar runs it and logs what it did. When it doesn't, it pages the right person with the full trail attached.
47 alerts received from checkout-api across Datadog, Sentry and CloudWatch.
Collapsed into one incident. Same root signature, three reporters.
Correlated with deploy v2.41.0, shipped 6 minutes earlier by @rhea.
Matched runbook db-connection-pool-exhausted at 0.94 confidence.
Ran remediation: scaled connection pool 40 → 80, drained two stuck workers.
Recovered in 31 seconds. Summary posted to #incidents. Nobody was woken.
Features
Write the fix once in plain English. Lunar recognises the signature next time and runs it, with every action logged for review.
Four tools screaming about one outage becomes one incident with four sources attached, not four separate pages at four separate minutes.
Every incident is checked against what shipped in the last hour, so the first question is answered before anyone asks it.
Schedules that respect time zones, holidays and the fact that people take leave. Handover notes are written for you.
If the primary doesn't ack in 90 seconds it moves on. No silent pages, no incident sitting unowned until morning.
Timeline, contributing factors and the actions taken, written up the moment an incident closes. You edit rather than start.
Integrations
Forty-plus native integrations, plus a generic webhook for the thing your team built in 2019 and never documented.
“We went from eleven pages a week to one. The thing I didn't expect is that people stopped dreading the rotation. Two engineers told me they'd have left over it.”
Pricing
Charging for alert volume punishes you for having a bad night. We don't do that.
For small teams and side projects that still deserve a night off.
Start freeEverything that makes the 3am page stop happening.
Start 14-day trialFor estates with compliance officers and a lot of services.
Talk to usQuestions
Nothing is suppressed silently. Every decision is written to the incident log with the confidence score and the runbook that matched. You set the confidence floor for auto-resolution, and anything below it pages a human by default. Most teams start at 0.95 and relax it once they have seen a month of decisions.
No. Plenty of teams run Lunar in front of an existing paging tool for the first few months, so Lunar does the triage and hands the survivors over. When you trust it, you can let Lunar do the paging too and switch the old tool off.
The ingest side is a URL and a token per source, so under ten minutes. Runbook automation is worth more time: teams usually import three or four existing runbooks in the first week, then add one each time an incident repeats.
Only the ones you explicitly grant. Actions are scoped with your own IAM roles, every run is logged and reversible, and you can set any runbook to propose-only so it writes out what it would do and waits for a human to approve.
Alert content stays in your region and is never used to train shared models. Enterprise plans can keep ingest entirely inside your own VPC. Lunar is SOC 2 Type II and the report is available under NDA.
Free for 5 responders · No card required · Live in under 10 minutes