Skip to content
lowkeylowkeyLowkey Docs

The agent's browser

Plenty of useful work lives on websites with no other way in: a supplier portal, a booking form, a dashboard behind a login. Agents handle these with a real Chrome browser running on your Lowkey box. It clicks, types and reads pages the way a person would, and you can watch it happen and step in whenever you like.

This browser runs on the Lowkey box, not on your own computer. To have an agent use the browser on your laptop or desktop instead, pair that machine; see Using your own computers (devices).

Watching it work

Section titled “Watching it work”

When an agent starts browsing for you, it opens a live view of the browser in the focus pane beside the chat. You see the real page, updating as the agent moves through it. The view has a few handles:

  • Fit view and 1:1 view switch between fitting the page to the pane and showing it at full size.
  • Pan and Touch modes change what a drag or tap does, which helps on a phone.
  • Fullscreen gives the page the whole screen.

The live view is view only while the agent is working, and a note says the agent has control.

Taking control

Section titled “Taking control”

Some steps need you: a sign-in, a two-factor code, a captcha, a judgment call on a form. Tap Take control and the browser is yours to click and type in. The agent waits while you have it. When you’re done, tap Hand back and the agent carries on from where you left the page.

Signing in this way means you type your password into the site yourself; it never has to pass through the chat. Agents will usually ask you to take over when they hit a login, and wait with you while you do it.

Sign-ins that stick

Section titled “Sign-ins that stick”

The browser keeps its own profile, the same way your everyday Chrome does, so a site you signed into last week is usually still signed in. You sign in once and later jobs on that site just work, until the site itself logs you out.

One browser per job

Section titled “One browser per job”

Each job gets its own browser workspace with a name, so two agents working at once don’t trip over each other’s tabs. When the job finishes the workspace is closed; an old live-view link then says “Browser workspace ended”. Ask the agent for a fresh look if you want to see the page again.

Pointing at something on the page

Section titled “Pointing at something on the page”

To show the agent exactly which button or field you mean, tap Freeze on the live view. It saves a still of the page and opens it for marking up. See Marking up and sharing your screen.

Section titled “What to read next”