Most monitors only fire when a pipe bursts. Nines catches the slow drip too, before it eats your error budget.
A 30-day SLO at three nines gives you about 43 minutes of error budget. A bad afternoon can chew through most of that without ever tripping a "down" check, because no run of failures was long enough to count. The leak is steady, the dashboard reads green, and the budget is gone before anyone looks at it.
We watch your error budget two ways: are you burning it down right now, and are you bleeding it slowly all week. Two checkpoints have to agree before we wake you up, so a single rough five minutes won't page you, and a real burn will. Under the hood it's multi-window burn-rate alerting, the same approach Google's SRE Workbook recommends.
Both windows must cross threshold before paging. Live trajectory in the product.
Your monitor checks at whatever interval you set. The moment a region reports a failure, that monitor ramps to 10-second checks for that region, and stays there until 10 successes confirm recovery. By the time the alert fires you have the data to back it up, and there's no extra noise the rest of the day. Most tools at this price point check on a fixed interval and call it a day.
Your budget for "fast enough" is separate from your budget for "up". Nines tracks p95 latency per region against a target you set, and runs a separate error budget against that target. When responses get slow, the latency budget starts burning even if uptime checks look happy, and the same burn-rate alerting fires on it.
Per-region p95 against your latency target. Budget burn is tracked separately from uptime.
Checking from one region means a single bad ISP route is enough to fire a page. Checking from many regions and alerting if any one of them fails is even worse — you've multiplied the chance of a false positive by the number of regions. Nines treats a check as failed only when more regions report it down than not. Strict majority required.
1 of 5 fails → page fires
Every transient ISP issue becomes a page.
1 of 5 fails → no page
A real outage that's visible from most of the world still fires within a check or two.
The features above are why most teams pick Nines, but a monitoring tool also has to handle the standard stuff well. It does.
HTTP/HTTPS, TCP, ICMP ping, and UDP. As fast as 30-second checks on Business plans.
For cron jobs and scheduled tasks. Send a ping when your job runs; we alert if one doesn't arrive in time.
Full chain inspection: expiry, issuer, SAN coverage. Warned days before any cert expires.
Public status pages with 90-day uptime history, incident timeline, and your branding. Bring your own custom domain.
Auto-opened incidents, manual updates, post-mortem timeline. Posts to your status page automatically.
Email and webhook delivery, retry-safe.
A free tier that's actually useful, a paid tier under $10, and burn-rate alerting on the Business plan. No per-seat fees, no overage surprises.
For side projects and small services
No per-seat fees.
For startups running real services
No per-seat fees.
For teams that run on SLOs
Lifetime access, limited edition
25 of 25 remaining