Private beta · onboarding 10 engineering teams
Stop noisy alerts before they reach your on-call.
BlipLite sits between Alertmanager or Grafana and PagerDuty or incident.io. It groups repeated alerts, explains what fired and applies your paging policy before anyone is woken up.
Or see what it does with a night of alerts- Runs in shadow mode alongside your current setup
- AI can't silently drop a page
- Keeps PagerDuty or incident.io
Your monitoring
- Prometheus Alertmanager
- Grafana Alerting
- Any JSON webhook
Add one webhook receiver. Keep your existing routes.
BlipLite
Shadow mode first- 01 · GroupRepeats of one rule, service and environment become one group.
- 02 · ExplainA summary of what fired. AI can raise severity, never lower it.
- 03 · Apply your policyDeterministic rules you write decide page, notify or digest.
- 04 · Record whyEvery decision is kept with its reason.
Your incident platform
- PagerDuty
- incident.io
Schedules, escalations and incidents stay where they are.
A night of alerts
37 alerts overnight. 2 pages.
A crash-looping worker re-sent all night, a flapping node exporter, checkout errors across four pods and a slow search API. This is what a 24/7 policy makes of it, with the reason for every decision.
- 37alert events in
- 4groups after grouping
- 2pages to the on-call engineer
- 35events grouped, held or explained
- PageHighErrorRatecriticalcheckout-api · production · 9 events · 4 alertsCritical pages at any hour. 9 events, one page.
- PageSearchLatencyHighhighsearch-api · production · 5 events · 1 alertHigh pages at any hour. 5 events, one page.
- NotifyKubePodCrashLoopingmediumnotifications-worker · staging · 13 events · 1 alertMedium notifies the team channel. Not urgent enough to page. 13 events, one entry.
- Resolved itselfTargetDownmediumflappingnode-exporter · staging · 10 events · 1 alertFlapping: fired 5 times and each one cleared within 10 minutes. Nobody is paged.
Sample data with made-up services. Not a customer result.
Run the same analysis on your own alertsThe obvious question
Why not PagerDuty AIOps?
If you're all-in on PagerDuty and AIOps already works for you, you may not need BlipLite. It's built for teams who want the decision made before the incident platform, by rules they can read.
- Where it works
- PagerDuty AIOps · On events after they arrive in PagerDuty.
- BlipLite · Upstream, before your incident platform sees anything.
- Works with
- PagerDuty AIOps · PagerDuty services and event orchestrations.
- BlipLite · Alertmanager, Grafana and webhooks in; PagerDuty or incident.io out. Switch platforms without redoing your alert logic.
- Who decides to page
- PagerDuty AIOps · Grouping models that learn from your team's behaviour, plus rules you configure.
- BlipLite · A deterministic policy you write. AI explains and can only escalate.
- Trying it
- PagerDuty AIOps · A standalone add-on on Incident Management plans, priced by events ingested.
- BlipLite · Shadow mode on a copy of your alerts. Your paging is untouched until you choose.
They aren't either-or. BlipLite hands its pages to PagerDuty, so AIOps can still group what arrives. The difference is where the decision is made, and who can read it. PagerDuty details from its AIOps documentation, October 2026.
Safety
AI can raise an alert. It can't quietly drop one.
A missed page costs more than a noisy one. So BlipLite is built to fail towards paging, and the parts that matter are enforced in code, not left to a model.
AI can't suppress a page
AI can raise an alert's severity and explain it. Holding back a page only ever comes from rules you set.
Enforced in the decision step: a lower severity from the model is ignored.
Your policy decides
Paging is deterministic: the same alert under the same policy gets the same decision, every time.
Presets today (24/7, business hours, no paging, digest only), custom rules next.
Fails towards paging
If triage errors or is slow, the alert carries on at its original severity.
Hard 8-second timeout. On any failure the rules-only result is used.
Every decision explained
Each group shows what came in, what triage said, which rule decided and why.
A per-alert timeline: in, AI, out.
Shadow mode is enforced
Try it on real traffic while your existing routing keeps paging as it does today.
Live paging needs three switches, checked in code. Previews can never page.
Secrets stay out of the model
Labels and annotations that look like tokens, keys or passwords are redacted before triage.
Redaction runs before anything leaves BlipLite. The count is shown on the alert.
Try it
Send it one alert.
Pick an Alertmanager alert and a paging policy, and see what BlipLite would do. Or copy the curl command and send it from your terminal.
{
"version": "4",
"groupKey": "{}:{alertname=\"HighErrorRate\"}",
"truncatedAlerts": 0,
"status": "firing",
"receiver": "bliplite",
"groupLabels": {
"alertname": "HighErrorRate"
},
"commonLabels": {
"alertname": "HighErrorRate",
"service": "checkout-api",
"namespace": "shop",
"env": "production",
"severity": "critical"
},
"commonAnnotations": {
"summary": "5xx error rate above 5% for 10 minutes",
"runbook_url": "https://runbooks.example/high-error-rate"
},
"externalURL": "http://alertmanager.example:9093",
"alerts": [
{
"status": "firing",
"labels": {
"alertname": "HighErrorRate",
"service": "checkout-api",
"namespace": "shop",
"env": "production",
"severity": "critical"
},
"annotations": {
"summary": "5xx error rate above 5% for 10 minutes",
"runbook_url": "https://runbooks.example/high-error-rate"
},
"startsAt": "2026-10-04T02:13:00Z",
"endsAt": "0001-01-01T00:00:00Z",
"generatorURL": "http://prometheus.example:9090/graph",
"fingerprint": "a1b2c3d4e5f60718"
}
]
}What BlipLite would do
Pick an alert and a policy, then send it.
Prepared examples for now, not live AI. Nothing you send is stored.
Design partners
Looking for 10 teams with noisy on-call.
BlipLite is early, and it's being built with the teams who use it. You'll work directly with Zee, who is building it, not a sales team.
A good fit
- A platform, SRE or DevOps team of roughly 5 to 50 engineers
- Prometheus and Alertmanager or Grafana Alerting
- PagerDuty or incident.io, with a real on-call rotation
- More pages than you'd like, and some you've stopped trusting
What you get
- Early access, in shadow mode, on your own alert traffic
- Help setting up shadow mode, from the person who built it
- A report of what BlipLite would have grouped, paged, notified or held, and why
- A say in which integrations and policy rules come next
What we ask
- Point a copy of your alerts at BlipLite (one webhook receiver)
- A 30-minute call every couple of weeks
- Honest feedback, including when it gets something wrong
Request beta access
Just your work email for now. A few optional questions come next, and I reply to every request myself.
Not ready? Run the Alert Noise Report on your own alerts first.
Straight answers
What engineers ask first.
BlipLite is a policy and decision layer for alerts. It sits between your monitoring (Prometheus Alertmanager, Grafana Alerting or any JSON webhook) and your incident platform (PagerDuty or incident.io). It groups repeated alerts, explains what fired, applies your team's paging policy and records why it made each decision.