How it works

Production monitoring runs per repository. This page covers where it looks, what makes something worth filing, and what a finished day leaves behind.

One run a day

Coldtea starts a run for each repository that has production monitoring running, on that repository's working days. The run is pinned to your local calendar day.

Every day is a working day to begin with. See settings and permissions to take days off.

The run looks for work, files what is worth doing, and closes the day with a brief. You can watch it happen on the production monitoring home, where the day's filed tasks and a live activity feed sit beside the brief.

Where it looks

Production monitoring reads two kinds of source.

Your repository. Coldtea reads what changed recently and proposes follow-up work tied to those changes. Separately, it works through the codebase one area at a time, so parts of the app that nobody has touched lately still get attention. Work found this way cites a precise location, shaped like path/from/root.ts:42, so you can check the claim before you act on it.

Your production signals. Coldtea reads the observability and deploy providers you have connected, such as error tracking, analytics, and deployments. Work found this way cites the signals it came from and can carry an impact estimate: how many users are affected, an event count, a window, and whether the trend is rising, flat, or falling.

Production is optional. With nothing connected, Coldtea still plans from the code alone. See connect signals to add it.

What makes something worth filing

Coldtea does not invent work. Everything it files carries proof: an error, a failing test, a code reference, an analytics number, or a piece of feedback. Anything without evidence behind it does not become a task.

From what survives that bar, Coldtea drops duplicates, ranks the rest, and files the highest-ranked work until the day's capacity budget is spent. Each task is sized as it goes. See settings and permissions for how sizing and the budget work.

How it knows your repository

Before any of this is useful, Coldtea needs a picture of what your repository actually is. It builds one when you first set up production monitoring, refreshes it periodically, and can rebuild it on demand. Each refresh updates the existing picture rather than replacing it, so notes and history carry forward.

That picture covers the kind of repository you have, the areas inside it and what each one is for, and how well the code is set up to report server errors, client errors, and product analytics. Because it records which areas are critical, production monitoring can weight work toward the parts of the repository that matter most.

What a run leaves behind

A finished run leaves a record you can read: what it looked at and found, the tasks it filed, and the day's brief.

Any task it starts runs as a cloud run, so you can open the task and use View session, or Watch live while it is still going, to see exactly what the agent did. Review it the way you would review any agent work. Nothing merges without you.