Always on watch
Every alert source, 24×7. It acknowledges before your phone lights up.
AI SRE employee
It takes the 2:47 AM page. You sleep.
EdgeX11 Nightwatch watches your alerts around the clock, owns the incident when one fires, and does the RCA and five-whys before you're awake.
Root cause in minutes. Not hours.
No credit card · connect Alertmanager in minutes.
02:47
Alertmanager · checkout-service — error rate 43% Alert
02:47
Nightwatch acknowledged. Nobody was paged. Owned
02:48
MetricsLokiDatadogK8sCloud
02:49
Seen before: INC-2214, 3 months ago — OOMKilled, cache never evicts. Memory
02:53
Mitigated. Pods recycled, error rate normal. Resolved
07:30
Post-mortem in your inbox — RCA, five whys, the permanent fix. For you
Reads alerts, metrics and logs from
- Alertmanager
- Prometheus
- Grafana
- Loki
- Datadog
- PagerDuty
- Kubernetes
- AWS
- Azure
- GCP
Always on watch
Every source. All night.
What it does
Watches. Owns. Explains.
Correlates everything
Metrics, Loki logs, K8s state and your cloud — AWS, Azure or GCP — read together, in seconds.
Remembers every incident
Déjà vu in seconds — it recalls the fix from three months ago.
Five whys, written
Not just what broke — why, five levels deep, in a chain you can read.
Guardrailed remediation
Safe mitigations run unattended. Production changes wait for your approval.
Post-mortem by morning
Timeline, root cause and the permanent fix — delivered, and remembered.
Root-cause analysis
It doesn't stop at what broke.
Why?Checkout died — pods were OOMKilled.
Why?Memory grew unbounded all week.
Why?The cache never evicts entries.
Why?The TTL was dropped in release v2.31.
RootA library default changed silently — and nothing pinned it.
In minutes. Not hours. And it never forgets the answer.
One incident, start to finish
Paged to post-mortem, unattended.
It's paged
The alert fires. Nightwatch takes it.
It investigates
Metrics, logs, deploys — and every past incident.
It mitigates
Inside guardrails. Prod changes wait for you.
You read it
RCA, five whys, the fix — over coffee.
Human control. AI execution.
It never gambles with production.
Nightwatch handles
- Acknowledge & investigate
- Correlate with past incidents
- Safe, pre-approved mitigations
- Write the RCA & post-mortem
You approve
- Production rollbacks
- Config & infrastructure changes
- Anything it hasn't seen before
- The permanent fix
Minutesfrom page to root cause, not hours
Every actionlogged, with a global kill switch
Zero2:47 AM pages that could have waited
FAQ
Quick answers.
Where does it get alerts?
From the monitoring you already run: Alertmanager and Prometheus natively, plus Datadog, PagerDuty, Loki and plain webhooks — and it reads Kubernetes and your cloud (AWS, Azure or GCP) directly during an investigation. When an alert fires, Nightwatch acknowledges and investigates instead of paging you.
Does it fix things on its own?
Only inside the guardrails you set. Safe, reversible mitigations run unattended; production rollbacks, config edits and infra changes are prepared and wait for your approval. Everything is logged, and a kill switch stops it all at once.
What's in the post-mortem?
The incident timeline, the root-cause analysis, the written five-whys chain, what was mitigated, and the permanent fix it recommends. In your inbox by morning — and saved to memory forever.
How does it know past incidents?
It works from your organizational memory. Every incident links to the services, deploys and fixes around it — so when something fires that looks like three months ago, it knows in seconds.
Will it ever page me?
Only when a remediation needs your approval, or when something is genuinely new. Everything else becomes a morning read instead of a 2:47 AM page.
What does it cost?
It runs as a Reliability employee on your plan — priced per employee, never per seat. See pricing.