FlakyWatch watches your GitHub Actions, puts a dollar cost on every flaky test, and turns each one into a tracked, owned GitHub issue that won't get lost.
Free forever for 1 repo · No credit card required · 2 minute setup
The hidden tax on every pull request
You’re paying for every re-run
A flaky test triggers a re-run. That’s CI minutes plus ~30 minutes of context-switching per developer. A team of 10 burns $3,000–$8,000/month on this, and it never shows up on a ticket.
Nobody owns the fix
Flaky tests fall between teams. Without an owner and a tracked issue, the same test breaks the same PRs for months. Re-running is easier than investigating — until a real bug slips through.
Your dashboard says green. Your team doesn’t trust it.
When engineers stop believing CI, they merge on red. That’s the moment flaky tests stop being annoying and start becoming dangerous. The fix isn’t more tests — it’s visibility and ownership.
Features
Detection is commoditized. Ownership isn't. FlakyWatch puts every flaky test on someone's plate, with a price tag attached.
Every flaky test gets a dollar amount attached — CI minutes plus the developer time lost to context-switching. Share the report with your team (or your CFO). The top 5 most expensive tests are one click away.
Every flaky test becomes a GitHub issue with classification, impact, and recommended next steps. Close it and it stays closed — until the flake returns, when we reopen it with a comment. No duplicates. Nothing gets lost.
Every test is classified as stable, flaky, broken, or newly unstable — with confidence scores, not guesses. Different problems get different priorities, so your team fixes what matters first.
When a known flaky test fails a PR, we comment so your team knows it’s not their code. No more ‘is this real?’ Slack threads.
Blocked runs, main failures, affected branches, cost per test — so you can defend fix time with data.
Monday email with new unstable tests, cost trends, and open issues. No noise if nothing changed.
Honest comparison
Flaky test tools aren't interchangeable. If one of these fits you better, use it.
If you’re a team of 1–5 devs on private repos
→ Use Trunk’s free tier. It’s free up to 5 committers and genuinely good.
If you need test quarantine today
→ BuildPulse ships quarantine at $99/mo. We don’t — yet.
If you only want a cost report, nothing else
→ Kleore’s free tier gives you that. Their Pro tier starts at $149/mo.
If you want every flaky test turned into tracked, owned, reopenable GitHub work — with a cost attached
→ That's what we built.
Setup
No agents. No config files. Just GitHub Actions + JUnit XML.
Install the FlakyWatch GitHub App and select repos to monitor. That’s it — no agents, no config files.
Add one step to your workflow that uploads JUnit XML results as an artifact. Works with pytest, Jest, JUnit, RSpec, and any JUnit-compatible runner.
- name: Upload test results
if: always()
uses: actions/upload-artifact@v4
with:
name: test-results
path: test-results.xmlFlakyWatch builds test history, classifies instability, and surfaces what needs attention first — with a dollar amount attached.
Dashboard
Classification, trends, cost tracking, and issue lifecycle — all in one place.
Flaky
12
+2 this weekBroken
3
+1 this weekNewly Unstable
5
+3 this weekTotal Tracked
847
Est. Wasted
$766
Blocked Runs
18
Main Failures
7
Wasted CI Minutes
~142 min
Sample data. Real costs are calculated from your GitHub Actions usage and a configurable engineering hourly rate.
Pricing
Every plan includes cost tracking, classification, and GitHub issues.
For solo devs and side projects
For small teams that want ownership
For platform teams
Install FlakyWatch on one repo. Free forever. Your first cost report arrives within a day of your next CI run.
Install on one repo — free