Chrome Extension

The Chrome addon is the half of the product that works in the browser. It captures screens and snapshots for visual checks, runs checks right on the open page and executes tests. Without it the product stays a reference book: the rules exist, but there is nobody to apply them to a real page. Below are the installation, the panel and runner mode.

Why the addon is needed

Captures a screen — the page’s image, its address, title, window size and structure: headings, buttons, fields.
Runs checks — executes the rules on the real page and shows what passed and what did not.
Executes tests — in runner mode it takes a ready test from the queue and walks it through step by step.
Takes snapshots for visual checks — captures the whole page, a component or an element, plus text snapshots: the page structure, styles, the accessibility tree and the order of addresses.
Executes builds — in runner mode it receives a build from the queue and walks its steps, capture included.

The addon is private: it is not in the Chrome Web Store. That is deliberate — every project’s checks are its own, and there is no need to keep them on public display. The product prepares a folder with the pairing to your server already written in, and Chrome only asks you to connect that folder.

One-click install

In the product, open Settings → UI Testing and press “One-click install and pairing”.
The app creates a test-agent-addon folder in a directory of your choice. Your server’s address and access key are already inside — nothing to type in.
Copy the address chrome://extensions, open it in a new tab and turn on the Developer mode switch in the top right corner.
Click “Load unpacked” and select the created test-agent-addon folder.

As soon as the addon connects, the dialog shows the green “Addon connected” line instead of the step list — no separate connection test is needed. If the browser cannot write the folder itself, a ready archive downloads: unpack it and select the folder the same way.

Manual pairing is needed only in the rare case — the addon is already installed, and the server is a different one. Then open the addon’s settings page, enter the server address and the access key, and press “Test connection”: the reply should be the list of your projects.

The panel on the page

Click the addon’s icon on the browser toolbar — the panel opens on the current page: a small window over the content. A second click or the ✕ hides it. The window can be dragged by its header, and it remembers its place: next time it opens where you left it. This works only on regular sites — on the browser’s service pages the icon flashes red, because there is nothing to check there.

Which project was recognized — determined by the address of the open page. No need to pick a project by hand: the panel finds the right one by the base address.
Screen capture — one button. The snapshot is saved at once, and you can immediately write a comment in words.
Rules for this page — the list of applicable checks and the “Check this page” button: it runs them all and shows the result.
Recording and running actions — start recording, work on the site as usual, stop; then run what was recorded and see how it went.
Ask the agent — describe the task in words, and the agent writes the action itself.
Last captures — the project’s twenty most recent screens with a quick comment box and a link to the full list in the app.

If the site is not registered yet, the panel offers to create a project: the name is prefilled from the page address, and the folder address is checked for validity right away.

Snapshot for a visual test

The panel has a separate card — Visual test. What it saves is not a training screen but a check snapshot: later it will be compared against the baseline, once a build passes the same step.

Four kinds of picture can be captured: the page’s visible area, the entire full page (the product scrolls it and stitches the frames together itself), a component or a single element — for the last two you point at what exactly to capture. There are also text snapshots: the page structure, styles, the accessibility tree and the order of addresses. They catch what a picture does not show.

Before the shot the product calms the page down so the snapshot does not flicker: it freezes time, turns off animations, hides the blinking input caret, normalizes the scrollbar, waits for the network to go quiet, the page to stop changing and the fonts to load, and blocks third-party domains — ads and counters — for the duration of the capture. All these measures are on by default, and the fine tuning lives in the project.

When several frames are taken, the product says honestly whether the page is “stable” or “unstable”: a flickering page will not slip into a comparison silently — it stays in the report as evidence of the flicker.

The panel never gets into the shot

A snapshot always shows the page without the panel: for the moment of capture it hides and returns to its place right after. Recorded actions do not notice the panel either — dragging the window and clicking inside it never get into the scenario. The product never captures or records itself.

Runner mode

The addon has a settings page. Runner mode is enabled there — it is on by default, and it is worth turning off only when this is your personal browser, where running tests would be inconvenient. The executor name is set on the same page: by it you can see which computer executed a run.

Enabled mode means the addon asks the server every minute whether a ready build is waiting and, if so, takes it and walks its steps. A runner can be paused — drained: it finishes the current build and stops taking new ones, and claiming can be switched back on at any moment.

The same page shows when the addon last talked to the server. The line “connected N seconds ago” means everything is alive; when the count goes into minutes, the browser is closed, the computer is asleep or the server is unreachable — the line says which one it is.

A step with a script is executed by the server

When a test step is a script, it is executed not by the browser but by the server. The addon is left with one request: ask the server to execute this step and record the result. No snapshot is taken for such a step — there is nothing to photograph, it has no page; instead of a picture, the step’s line shows its text output. What that step is and why it is needed — in the Scripts & Extensibility section.

Honest limitations

The addon is private — it is not in the Chrome Web Store, installation goes from the folder as described above. The browser will not update it on its own.
After a product update the addon is refreshed by hand — press “One-click install and pairing” again (the folder is rewritten) and press the update button on the addon’s card at chrome://extensions. Without this, Chrome keeps working with the old version.
One run per browser — a test occupies a separate tab, and a second test in the same Chrome will not start until the first one finishes. For frequent schedules, keep a separate executor computer.
Only the addon takes a visual snapshot — without a connected Chrome there will be no visual builds: there is nothing to compare until someone captures the page.
A test made of scripts alone needs no addon — when all of a test’s steps are scripts, the server executes it: neither a browser nor the addon is needed for such a run.

Next: what a screen and a comment are — in the Screens & Comments section.

← Back to the documentation index