Try it for free

Agentic quality assurance

You build it. We test it.

Write your tests in plain English and let qAuto handle the rest. Works on both mobile and web apps.

002-apply-coupon.md

Apply a coupon code

  1. 1. Open the cart.
  2. 2. Tap "Checkout".
  3. 3. Enter SAVE10 in the coupon field.
  4. 4. Tap "Apply".

Expected: the total drops to $45.00.

Terminal — my-shop
$ qauto run 2
FAILED 002 Apply a coupon code 8 steps · 16 credits · 43 s
Test step 4: Tap "Apply".
Expected:The total drops to $45.00.
Observed:The total is still $50.00.
Report:qauto/results/002-apply-coupon/2026-09-29_14-02-10/report.md
$ ls qauto/results/002-apply-coupon/*
report.json report.md screenshots steps.md video
$
  • Works with Claude Code and Codex
  • Android emulators
  • macOS, Linux, and Windows
  • No test code to maintain

How it works

From a sentence to a verdict.

No selectors, no test IDs, and no scripts that break when a button moves.

1Describe it

Write the steps and what should happen, in a Markdown file in your repository. Or ask your coding agent to write it.

2qAuto runs it

On an Android emulator. At every step, qAuto looks at the screen and does what a person would do next.

3You get the verdict

Passed, failed, blocked, or unclear, with the screenshots, a video of the run, and a report you can read.

Made for coding agents

Your agent writes the feature. qAuto checks it works.

A failed test says which step went wrong and what qAuto saw instead, so your agent can fix the bug and run it again.

  • qauto init teaches Claude Code and Codex to write and run tests
  • --json gives them one result per test
  • report.md for you, report.json for tools

Private by design

Your app’s screens aren’t kept.

qAuto needs to see your app to test it. It doesn’t need to keep what it saw.

Secrets stay on your machine

Tests use placeholders such as {{secrets.TEST_PASSWORD}}. The real values are filled in locally and never sent.

Screenshots aren’t stored

We keep only a fingerprint of each screen, which can’t be turned back into the image.

No training on your app

Screenshots go only to AI providers that never train on them. Zero data retention runs cost no extra, for now.

Open-source client

The command-line tool is Apache-2.0, published with the first release. See exactly what runs on your machine.

Pricing

Start free. Pay for the steps you use.

2 credits a step, and a test case is about 10 steps. Steps that fail because of us are free.

Free

100 credits

When you sign up, with no card.

About 5 test cases

Starter

US$5 a month

5,000 credits a month.

About 250 test cases a month

Builder20% more

US$10 a month

12,000 credits a month.

About 600 test cases a month

Unused credits roll over, up to a month’s worth. Plus tax where it applies.
See pricing

Stop clicking through your app by hand.

qAuto is in private alpha. Email us to join, and we’ll send you the client and your first credits.