What is QA Agent?

Code changes land fast. Your QA process shouldn't be the bottleneck.

What you'll learn

• What QA Agent does and why markdown test specs beat code-based tests
• How it fits into a real product workflow
• When to reach for it versus Playwright or Cypress

The core idea

QA Agent is an autonomous browser testing runner. You write test cases in plain markdown. QA Agent reads them, opens a real browser, executes the steps using an AI model (Gemini 3.1 Flash Lite by default), captures screenshots and screen recordings, and produces an HTML report with pass/fail results and visual evidence.

The key design choice: test specs are written in product language, not code. A spec looks like this:

# Checkout flow completes successfully

- **persona**: free-user
- **priority**: high
- **tags**: checkout, critical

## Steps

1. Navigate to /cart
2. Add the first product to the cart
3. Click "Proceed to checkout"
4. Verify the order summary is visible

## Expected

- Cart shows correct item count
- Order summary displays correct totals
- No console errors during the flow

This means a non-engineer can read and review a test case. More importantly, the spec survives a UI refactor. If a button's label changes from "Proceed to checkout" to "Review order", you update one line of markdown, not a selector chain spread across multiple test files.

How it runs

QA Agent uses browser-use under the hood, which drives a real Chromium instance. The AI model interprets each step in the spec and translates it into browser actions. Clicks, navigation, form fills, assertions. At the end of every run it generates:

• An HTML report with per-test pass/fail, timing, and failure notes
• Screenshots at each step
• GIFs and videos of the full browser session
• Structured failure notes with the exact step, reason, and a screenshot path

When to use QA Agent

Use it when:

• You want to verify user-facing flows before a release without writing and maintaining Playwright suites
• Your team wants QA coverage that non-engineers can read and expand
• You need visual evidence of failures, not just a red/green result

Use Playwright or Cypress instead when:

• You need sub-second CI performance (QA Agent runs at human browser speed)
• You need to test deeply technical internal APIs or WebSocket behaviour
• You already have a mature, maintained selector-based suite

What's next

Cable 2: Setting Up QA Agent →