Where Crumbtrail fits.
Crumbtrail captures the session, the network calls, the active flags, and the row-level database state at the exact moment a bug fires — automatically, on every failure. That turns three familiar dead ends into a lookup. Here's what changes for the people who hit them.
When your agent picks up the ticket
Your coding agent reads the ticket text and your repository — and nothing about the run that actually failed. So it patches what the words describe, not what the machine did. Hand it the captured session and it fixes the real defect on the first pass.
Settings that save but don't stick
The ticket says “my changes don't save.” The agent reads the form component, tightens the validation, and ships a fix. The real bug — a PATCH that returned 200 while writing to the wrong tenant scope — is untouched, and the ticket reopens a day later.
The agent gets the captured request and the row-level database diff showing the write landed on the wrong tenant. It corrects the scope check and closes the ticket in one pass.
When the pager goes off at 2am
An error-rate spike doesn't arrive with reproduction steps. Without the failing run in front of you, the night is a guessing game against production. With it, root cause is a lookup.
Checkout error-rate spike, 2am
Error rate on checkout jumps. You restore a multi-terabyte tenant backup to a staging box, guess at the sequence that triggers it, and page the customer for steps they don't remember. Two hours later you have a maybe-repro.
You open the exact failing session: the request that broke, the flags that were live, and the database state at the moment it fired. Root cause in minutes — no restore, no repro.
When support escalates a bug
A screenshot and a customer's retelling isn't a reproduction — it's the opening move in a back-and-forth. Attach the captured session and the escalation lands reproducible on the first try.
“The dashboard is blank”
Support files “dashboard is blank,” attaches a screenshot, and asks the customer to clear their cache. Engineering can't reproduce it, asks for the browser and the steps, and four round-trips later still has a guess.
Support attaches the captured session — the failed API call, the console error, the exact state. Engineering opens a ticket that reproduces on the first read and goes straight to the fix.
See where it fits.
Embed Crumbtrail and your next bug arrives with the session, the request, and the database state already attached — for your agent, your on-call, and your support team. Or tell us what's breaking and we'll show you what it would have caught.