QAssistWrite a goal. Read a verdict.

Browser testing with no selectors and no scripts. Describe what should happen in plain English; an AI agent drives a real Chromium, streams it live, and decides whether it happened.

Your first run

Watch a real run

Run it yourself

No selectors to maintain

A test is a sentence, not a script. When the button moves or the class name changes, the agent still finds it — there is nothing to update.

You watch it happen

The browser session streams to the page while the run works, at about six frames a second, and is recorded so you can watch it back later.

A verdict with reasons

Every run ends passed or failed with a written rationale, the step list the agent took, and the failed requests and console errors the page produced.

Yours to run

Self-hosting is free, for anything, forever. Model tokens are bring-your-own on every tier, so the only bill is the one you already have.

What this manual covers

This is the manual for using QAssist. It follows the path a test actually takes: one goal typed into a box, then the same goal saved, grouped, given credentials, scheduled, and finally gating a deploy.

New here? Your first run gets you a verdict in about five minutes, whether you are on a hosted instance or your own box.
A run keeps failing and you think it shouldn't?Writing a goal is almost always the answer, and When a run goes wrong covers the rest.
Running your own instance? Self-hosting and Settings.

Notes on the internals — schema rules, the deployment chain, the design system, the test philosophy — are not here. They live in docs/ in the repository and are written for someone editing the code.