Features / Capture & fix
Error monitoring intake
Start before anyone complains
Connect your error tracker and Tore starts work on a new error the moment it appears, often before the first customer notices, and usually before anyone writes in.
- 10:41:58 POST /api/checkout 500
- 10:41:59 TypeError: priceId undefined
- 10:42:03 POST /api/checkout 500
- 10:42:07 POST /api/checkout 500
- 10:42:11 TypeError: priceId undefined
Reading your repository
- app/(main)/checkout/page.tsx
- api/checkout/route.ts
- api/checkout/session.ts
- lib/stripe/client.ts
Proposed change
Waiting for the investigation to find the frame…
What it does
- New errors become investigations without a human filing anything
- Grouped by fingerprint, so one bad deploy is one investigation
- Thresholds decide what is worth picking up and what is just noise
- Rules for what to always ignore and what to always escalate
Without it
The error was in your logs for a week before a customer wrote in
You find out from a customer
Something broke at 09:14. You hear about it at 11:40, from someone who is already annoyed, in a message that says 'is checkout down?', and by then it has hit forty other people who just left.
The alert channel is noise
Your error tracker fires into a channel nobody reads, because the last two hundred alerts were the same known issue. The one that mattered scrolled past at 2am between two of them.
Knowing is not fixing
Even when you see it, the alert is a stack trace with no owner. Someone still has to stop what they are doing, reproduce it, find the cause, and write the fix.
How it works
Connect the tracker you already run
Sentry today; Rollbar, Bugsnag and Datadog are on the way. Read-only, and you can disconnect it in a click.
What that includes
- Sentry today; Rollbar
- Bugsnag
- Datadog are on the way
Noise is grouped before anything acts
Ten thousand events from one bad deploy collapse into one fingerprint. Thresholds decide what is worth picking up; an ignore list holds the things you already know about.
An investigation opens on its own
No human files anything. Tore pulls the release, the commit, the stack frame and the surrounding code, and starts working while the errors are still arriving.
What that includes
- Tore pulls the release
- the commit
- the stack frame
- the surrounding code
It stops at the gate
A change is prepared and waits. Anything touching a locked risk class, authentication, billing, migrations, personal data, escalates to a person instead of proposing.
What that includes
- authentication
- migrations
- personal data
What it means for you
The fix often lands before the first ticket
The best support conversation is the one that never starts. When it does start, the answer is already 'fixed, shipping today' rather than 'can you tell me more?'
Your engineers stop being the alert triage rota
The repetitive part, reproduce, locate, understand, is done before anyone is interrupted. What reaches a person is a decision, not a search.
The people who reported it get told
Because the tracker and the support queue are the same system, closing the error closes the conversation too.
The honest limit: Opening the pull request on your repository is not switched on yet, the change is prepared and waits for your engineer.