Feature · Development · Automation
Your tests,
in your terminal.
qamind-cli puts the whole loop where you already work: describe a workflow in plain English, read the plan it drafts, approve it, and run it in a real browser — then wire the same commands into CI.
In your terminal
Describe it. Review it. Run it.
One install, no project scaffolding and nothing to configure. qamind init picks the workspace and project once, and every command after that knows where it is.
1$ npm install -g qamind-cli2$ qamind init3$ qamind test create "Sign in and confirm the dashboard greets the user"5Plan — 4 steps · 3 assertions71 Open /login82 Fill Email with $USER_EMAIL93 Fill Password with $USER_PASSWORD104 Click Sign in11Assertions131 URL matches /dashboard you asked for this142 Welcome message is visible you asked for this153 Avatar menu is visible suggested16accept 3 · approve · refine <text> · run · export18$ qamind suite run "Checkout" --reporter junit2012 passed · 1 failed · qamind-junit.xml written
1# .github/workflows/qa.yml2- run: npm install -g qamind-cli3- run: qamind suite run "Regression" --reporter junit --out results.xml4env:5QAMIND_TOKEN: ${{ secrets.QAMIND_TOKEN }}6Regression — 24 cases8✓ 22 passed ✗ 1 failed — 1 skipped9results.xml written · exit 110$ qamind flaky --since 7d123 tests failed then passed with no change:13checkout-guest 4 flips14search-filters 2 flips15profile-avatar 2 flips
In the product
How it actually works.
Install and point it somewhere
npm install -g qamind-cli, then qamind login and qamind init. Init asks which workspace, project, environment and URL to use, and which machine runs the browser. Nothing is assumed, and a bare word is refused rather than stored as a host.
Author in English, approve what ships
qamind test create "…" drafts a plan you read — steps and assertions in plain language. Assertions the agent invented are listed as suggestions; accept the ones you want. Nothing runs until you approve, and the agent cannot approve its own work.
Run it, watch it, export it
qamind test run and qamind suite run execute in a real browser and stream progress; run watch follows one, run inspect shows what happened, and test export writes Playwright code you can keep.
Put it in the pipeline
The same commands run unattended with a QAMIND_TOKEN. A failing suite exits non-zero, --reporter junit writes XML your pipeline already reads, and qamind flaky names the tests that pass and fail without a code change.
See these steps on screen
coming soonA guided video of this exact flow is being recorded — it will live right here.
Spec sheet