qaly

QA evidence for agent-written code

Your agent says it fixed it.
This is what actually happened.

Qaly records a real session in the browser — every click, every thrown error, every failed request — and hands the whole trail back to the agent that wrote the code.

Get the recorder See the loop
run 4f2a · checkout regression
A recorded run: the tester reaches checkout, submits, and the page throws. Qaly keeps the click that caused it, the stack, the failed request and the tester's annotated screenshot as one finding.

One loop, closed

Most bug reports lose the evidence between finding the problem and fixing it. Qaly keeps it attached the whole way round.

  1. 01

    The agent asks for a test

    It finishes a feature and registers a task: what changed, what to check, which URL.

  2. 02

    A person records

    The tester opens the side panel, lands on the right page, and presses record.

  3. 03

    Qaly captures

    Clicks, thrown errors, failed requests and route changes, in order, with timing.

  4. 04

    A finding is frozen

    The tester reports it. That slice of evidence stops changing, so it stays readable later.

  5. 05

    The agent reads it back

    Over MCP it gets the timeline, the stack, the failed calls and the screenshot.

  6. 06

    A person confirms

    The fix raises a retest. Only a human verdict marks it verified — a fix cannot confirm itself.

What it captures

  • Clicks and form input, with a durable description of the element
  • Errors thrown by the page, with the stack as the page saw it
  • Failed requests, including the status when the stack is gone
  • Route changes in single-page apps, not just full page loads
  • Screenshots of the tab or the whole screen, croppable and annotated

What it will not do

  • Record a tab whose origin you have not listed against a project
  • Store the contents of a password field — only that one was filled, and its length
  • Let an agent mark its own fix verified
  • Send anything to a third party. It talks to Qaly and nothing else

Recording every tab would sweep a tester's email and banking into evidence an agent later reads. Scope is by origin on purpose.

Get the recorder

Qaly records from a Chrome extension. Download it, load it, sign in.

Download for Chrome

  1. Unzip it

    Anywhere you will not delete by accident — Chrome loads it from that folder.

  2. Load it

    Open chrome://extensions, turn on Developer mode, choose Load unpacked.

  3. Sign in

    Press Sign in in the extension's options. No token to copy.

Access is by invitation.