Setting up an environment on the right branch to repro is slow and flaky.
Generate a test plan, execute it on the real app, and gate every PR.
Turn a feature into a test plan you review, run it against a live sandbox with video proof, and wire it up so every pull request runs the same checks — failing tests are fed back and fixed automatically.
One real thread, start to finish
Generate a test plan for checkout with coupon codes — happy path plus the edge cases that bite us.
- Valid coupon applies the discount to the total
- Expired coupon is rejected with a clear error
- Stacking two coupons is prevented
- Coupon + tax rounds correctly
Looks complete. Run it on the live sandbox and capture anything that fails.
- Valid coupon applies the discount to the total
- Expired coupon is rejected with a clear error
- Stacking two coupons is prevented
- Coupon + tax rounds correctly
Is that last one a real bug or expected rounding?
Real bug — the discount is applied after tax, so the total is a cent high:
- Cause — rounding happens before the coupon is subtracted
- Impact — every coupon order with tax is off by rounding
References
Sam wires the plan in as a PR test gate, then opens the fix.
Generate a test plan for checkout with coupon codes — happy path plus the edge cases that bite us.
- Valid coupon applies the discount to the total
- Expired coupon is rejected with a clear error
- Stacking two coupons is prevented
- Coupon + tax rounds correctly
The test gate catches it and feeds the failure back.
Once the plan is wired as a gate, it runs on each PR. A failing case is sent back to be fixed and re-run — no manual smoke test required.
Test gate ran the coupon plan — 1 case failed
Gently moved rounding after the discount and pushed the fix
Test gate re-ran — 14/14 passed
Today this plays out as separate follow-up runs in the thread.
Test plans live in a doc; nobody executes them consistently.
Manual smoke tests don’t run on every PR, so regressions slip through.
QA & testers, in one workspace.
Generate test plans
A plan built from the real feature — happy path plus the edge cases.
Execute on the real app
Run the plan on a live sandbox and capture screenshots + video of failures.
Bug or expected?
Settle it from the real code instead of escalating to engineering.
PR test gate
Every PR runs the plan; failing tests are fed back and fixed automatically.
Give qa & testers a safe way to work with the product.
Start with questions. Grow into plans, previews, tests, and pull requests when your team is ready.
Request access