AI Root Cause
A likely cause in seconds, with the evidence.
When a monitor goes down, Xuptify looks at what changed around that moment: deploys, config changes, error rates and resource metrics. It shows the most likely cause, the data behind it and how confident it is. Your engineers decide what is true and what to do next.
- Deploy v4.18.2 by ci-bot at 20:39:04, 2 min 8 s before first failure
- DB connections 48 → 200 / 200 (100%) from 20:40:51
- New error PoolTimeoutError · 3,912 events in 60 s
- Failing in all regions, so not a network or CDN issue
Ranked hypotheses with supporting evidence
How it works
Set up in four beats.
- 01
Connect read-only sources
Deploy events from GitHub, GitLab or CI, errors from Sentry, metrics from Prometheus, Datadog or CloudWatch.
- 02
Correlate the window
Changes in the minutes before the failure are lined up against the exact time checks began failing.
- 03
Rank the hypotheses
Each possible cause gets a confidence level and the specific evidence that supports it.
- 04
Engineer confirms
Mark a cause as confirmed or wrong. The decision is recorded on the incident timeline.
Capabilities
What you get with AI Root Cause.
Deploy correlation
Which release went out, when, by whom, and how close it was to the first failed check.
Error spikes
New error types and sudden rate changes from your error tracker.
Metric shifts
Database connections, CPU, memory, queue depth and latency changes.
Regional patterns
Is it every region or only one? Separates your bug from a network problem.
Evidence first
Every suggestion links to the exact deploy, error group or graph it came from.
Human confirmation
Nothing is marked as the cause until an engineer agrees.
Specifications
The details engineers ask about.
How we use AI responsibly
- AI Root Cause suggests likely causes with evidence. Engineers confirm and decide what to do.
- Xuptify never changes your systems. It only reads the data you connect.
- Integrations use read-only access where possible, and API keys are encrypted at rest.
- Your incident data is used to serve your account only. It is encrypted and never sold.
- Deploy sources
- GitHub, GitLab, Bitbucket, CI webhooks
- Error sources
- Sentry, webhook ingestion
- Metric sources
- Prometheus, Datadog, AWS CloudWatch, Grafana
- Access
- Read-only wherever the provider supports it
- Output
- Ranked causes, confidence, linked evidence
- Availability
- Team and Scale plans
FAQ
AI Root Cause questions
Does AI fix things automatically?
No. It suggests the likely cause and shows why. Rolling back, restarting or changing anything is always your team's decision.
What if the suggestion is wrong?
Mark it as wrong and add the real cause. That is recorded on the timeline and in the postmortem draft.
Does it need write access to our tools?
No. Read-only access is enough, and that is what we ask for wherever the provider allows it.
Same pulse
Works together with
60 bpm 10 monitors · 1-minute checks · free, no time limit