Watch · Causal trail

Causal trail

Your alerts point at the line of code that caused them. Imports, callgraph, schema/route reads + writes joined to Sentry, Vercel, PostHog, and DB slow queries - a cross-signal correlator wires each alert to the change that likely triggered it.

Where to find it

  • Localhost: Build → Causal trail · the top item every time prod alerts
  • API: GET /api/incident-attribution?repo=<id>&event=<prod.error-id>
  • MCP tool: get_incident_attribution - pass an event, get back its confidence-banded attribution chain (the introducing deploy, its commit, and the session)
  • Keyboard: ⌘ K then trail

What it does for you

One graph, four signals.Sentry crashes, Vercel deploy fails, PostHog event regressions, and DB slow queries all flow into the same causal graph, joined to the file or schema column they touch.
Each alert names its likely cause.The correlator matches each alert to the deploy that shipped just before it (by commit ref, or the nearest preceding deploy within a 30-minute window), then blames the culprit line back to the session that wrote it. The top suspect lands on your dashboard with the diff one click away.
Opt-in per source, off by default.Sentry, Vercel, PostHog, and DB-slow-query ingest are each opt-in per repo with redaction at ingest - nothing leaves your machine until you turn it on.

Read more