One of the most annoying parts of working with CI isn’t failures. It’s uncertain failures.
A test fails. You rerun it. It passes. You rerun again. It fails.
Now you're stuck in a loop:
1. rerun workflow
2. open GitHub Actions
3. wait
4. check result
5. repeat
Not because you enjoy it, but because you're trying to answer a simple question: “Is this test actually broken, or just flaky?”
This loop quietly destroys focus. You keep context-switching while waiting for runs to finish, and it adds up.
I built a small tool to deal with this: SubCat.
It’s a macOS app that lets you monitor your Pull Requests or paste a GitHub Actions run URL and then forget about it. When the run finishes, you get a native notification (success/failure) and a click takes you straight to the result.
But the main reason I built it is repeat mode.
You can tell SubCat to rerun the same workflow N times and come back with a report in Markdown format.
It’s not trying to be a full observability platform like BuildPulse or Trunk.
There’s no setup, no workflow changes, no SaaS.
It’s just a lightweight, local tool for when a flaky test shows up and you need an answer now.
SubCat is:
- open source
- free
- macOS-only (for now).
One of the most annoying parts of working with CI isn’t failures. It’s uncertain failures.
A test fails. You rerun it. It passes. You rerun again. It fails.
Now you're stuck in a loop: 1. rerun workflow 2. open GitHub Actions 3. wait 4. check result 5. repeat
Not because you enjoy it, but because you're trying to answer a simple question: “Is this test actually broken, or just flaky?”
This loop quietly destroys focus. You keep context-switching while waiting for runs to finish, and it adds up.
I built a small tool to deal with this: SubCat.
It’s a macOS app that lets you monitor your Pull Requests or paste a GitHub Actions run URL and then forget about it. When the run finishes, you get a native notification (success/failure) and a click takes you straight to the result.
But the main reason I built it is repeat mode.
You can tell SubCat to rerun the same workflow N times and come back with a report in Markdown format.
It’s not trying to be a full observability platform like BuildPulse or Trunk. There’s no setup, no workflow changes, no SaaS.
It’s just a lightweight, local tool for when a flaky test shows up and you need an answer now.
SubCat is: - open source - free - macOS-only (for now).
Repo: https://github.com/semisse/subcat
Curious if others deal with flaky tests in a similar way, or if you’ve found better approaches.