Skip to content

Live browser

Live browser shows what the roles that currently have a browser open are seeing. It is not a video: it is a snapshot that refreshes every so often. At 700 ms you can see perfectly well what a role is doing, and asking for per-role video would multiply the cost for nothing.

You get there from Settings, Browser tab; the top bar reads “Live browser”.

Only roles with the Navigate skill appear. A role without that skill never shows up here, because it does not browse.

Situation What you see
Nobody browsing “Nobody is browsing right now”, with the explanation that when a role with the Navigate skill enters a site, you will see its screen, the step it is taking and from which address
It cannot be queried “The browser could not be queried. There is no browser running: check that the role has the Navigate skill and that the office is still responding.”

At the top, a small bar with two things:

Element What it does
“N roles browsing” The count. In the singular, “1 role browsing”
Columns 1 / 2 / 3 Changes how many screens fit per row. The chosen one stays marked and is remembered between visits. Two by default

Each screen is a card:

Element What it is
Green dot and role name On the left of the header. The name is a button: it opens that screen large (⤢)
Action On the right of the header: what it is doing right now
Image The page capture, refreshed on its own
Text box “Type on the page and press Enter…”. Types into that role’s browser
Footer The page title; if there is none, the address. Hover to read the full address

The grid is for watching several at once, but at 320 px you cannot read a page. That is what the expanded view is for.

The text next to the name is written by the role itself while it browses:

Action What is happening
opening It has just landed on a page
reading It is pulling the content out of the page
step N of M It is on step N of a sequence of M
waiting It is waiting for something from the page
waiting for you to resolve it It has run into something it cannot do alone and is waiting for you
waiting for you to resolve it (N s) The same, with the time it has left: it holds for up to 300 seconds
captcha solved, continuing It is unblocked and carrying on
you clicked / you typed The last intervention was yours

You click directly on the image. The cursor turns into a crosshair and, while the click is being sent, into a progress bar with the image dimmed.

What it is for: the things an agent cannot do alone. Passing a captcha, accepting the cookie banner that gets in the way or picking from an odd dropdown.

Action What it does
Clicking on the image Sends a click to that exact point on the role’s page
Typing and pressing Enter Sends the text to the role’s browser, wherever the focus is

Two warnings worth being clear about:

  • The click is proportional to the size of the image, so it does not matter how large you see it: it lands where you clicked.
  • The text is typed, not submitted: Enter is not pressed on the page. If a form needs submitting, you then have to press its button. The server cuts the text at 500 characters.

If something fails, a notice box appears over the grid: “Could not click.” or “Could not type: the browser is not responding”, with the reason the server returns, if there is one.

Stepping in is stepping in for real: you are touching that role’s browser, not a copy.

Clicking a role’s name (or the ⤢) replaces the grid with that screen alone, at whatever width there is. It is where you actually read.

Element What it does
← All Goes back to the grid
Name and action The role and what it is doing
Title or address On the right, truncated. Hover for the full address
Large image Same as in the grid: click to step in
Text box Same as in the grid
Footer “Click on the page to step in. Escape returns to all.”

Escape also returns to the grid, from any point in the expanded view.

  • The list of browsing roles is queried every 1.2 seconds.
  • Each image is requested again every 700 ms.
  • A screen that has given no sign of life for more than a minute stops appearing: if the role closed the browser or hung, it does not stay there as if it were still working.
  • The list comes sorted by role name.

If a role’s browser closes, its screen simply disappears from here. There is no history: this view is the present, not a record. What it did stays in its report and on the board.