Agentic QA
for web apps

Give it a URL and a plain-English user story. It drives a real browser with an LLM, records video, and has a second model independently verify the result.

Get Started View on GitHub

terminal

pip install qaprobe

qaprobe install

qaprobe run --url https://example.com \

--story "Search for 'test' and verify results appear"

✓ Verdict: PASS | a11y: 0 violations | Video: runs/report.html

How it works

Three steps. Zero boilerplate.

Describe

Write a plain-English user story. No selectors, no test IDs, no code.

Execute

QAProbe drives a real Chromium browser using the accessibility tree. Every action is recorded.

Verify

A second, independent model reviews the result. Three-way verdict: PASS, FAIL, or INCONCLUSIVE.

Features

Everything you need.
Nothing you don't.

Real Browser Testing

Playwright-powered Chromium. Video recording and trace on every run.

Accessibility-First

Uses the AX tree, not CSS selectors. Free a11y audit on every run.

Independent Verification

Two-model architecture: one acts, another verifies. No self-reported results.

Plain English Stories

No code, no selectors. Describe what a user would do.

CI Ready

GitHub Action included. YAML suites for organized test runs.

Open Source

MIT licensed. Run it yourself, extend it, contribute.

Quick start

Up and running in
under a minute.

terminal

$ pip install qaprobe

$ qaprobe install

$ qaprobe run --url https://example.com \