Find friction before a ticket is filed.
Deterministic detectors run on every session as it is processed.
Issues are clustered by type, selector, route, and error fingerprint, ranked by impact, and routed to Slack, Teams, email, PagerDuty, Opsgenie, Jira, Linear, GitHub, Azure DevOps, or a webhook.
Signals 0 fired
Why generated from this session
In this session Signals fires an error click, a rage click, a dead click, and form abandonment, each pointing at its moment in the replay.
Rules first, AI second.
Twenty-two deterministic detectors run first. Rules are cheap, explainable, and give the AI grounded evidence. Thresholds are tunable per project.
| Signal | Rule (defaults) | Severity |
|---|---|---|
| Rage click | ≥ 3 clicks within 1 s on the same element or within 30 px | Medium; High if followed by exit |
| Dead click | Click on an interactive-looking element with no DOM change, navigation, or request within 1 s | Medium |
| Error click | Click followed within 500 ms by a JS error or a failed request it initiated | High |
| Thrashed cursor | Fast, high-entropy pointer movement with no clicks for over 3 s | Low |
| Form abandonment | A field changed, no submit, page left or 60 s idle | Medium; High on checkout and intake |
| Field struggle | Same field edited three or more times, or a validation message appears | Medium |
| Failed request | 4xx or 5xx, or a network error, excluding allowlisted probes | High for 5xx |
| Slow request | Over the endpoint's 7-day p95, or over 3 s | Low or Medium |
| Frozen UI | No paint or mutation for 2 s after an interaction, or a long task over 1 s | High |
| Layout instability | CLS over 0.25 on a page, or a shift right before a click | Medium |
| Slow page | LCP over 4 s or INP over 500 ms | Medium |
| Navigation loop | A → B → A → B within 30 s | Medium |
| Blank page | Almost no visible text 3 s after load | High |
| Console error burst | More than 10 console errors in 5 s | Medium |
| Crash | Session ends without pagehide and the next load finds unflushed data | High |
| Accessibility trap | A keyboard user cannot leave an element for 10 s | Medium |
| Monitor trigger | Custom trigger on a kiosk or telehealth visit, with pre-roll | High |
| Custom | Your own rules on custom events and properties | Configurable |
Also detected: back-button bounce, broken resources, offline periods, consent friction. See the docs for every threshold.
One issue, not a thousand alerts.
- Detect per session, as chunks arrive
- Cluster kind + selector + route + fingerprint
- Rank count, users, trend, release
- Explain AI names it, cites evidence
- Route Slack, Teams, PagerDuty, Opsgenie, Jira, Linear, GitHub, Azure DevOps, webhook
Regression alerts
A cluster that exceeds three times its 7-day hourly baseline fires once, with counts updated in place on Slack or Teams rather than re-posting.
Release attribution
Every issue carries the release it was seen on. New clusters after a deploy are the first thing the digest mentions.
Lifecycle
Acknowledge, snooze, assign, resolve, or mute from the notification itself. Closing the Jira, Linear, GitHub, or Azure DevOps item resolves it.
Product updates by email.
Leave a work email and we will tell you when something ships. Or book a demo with an engineer.
You are on the list.
Product updates will go to .
Free for 1,000 stored sessions a month, with unlimited seats. Create a workspace and add one script tag.
Start