VeilBrowserDocs

Cookie warm-up

Let a new profile visit ordinary sites before its first real use, so it arrives with a browsing history and cookies.

A brand-new profile has no cookies, no history and no storage, which is itself a signal some sites look for. The warm-up robot gives it a past: it opens the profile without a window, visits a list of sites, scrolls each one for a while like a reader, and closes it again. Everything the visits leave behind is saved with the profile's browser data.

Warm-ups run on a computer with the desktop agent, through the profile's own fingerprint and proxy.

From the dashboard

  1. Open the profile's details from the profiles list and choose Warm up.
  2. List the sites, one per line, or choose Use popular sites for a starter list.
  3. Pick how long to stay on each site (3 to 120 seconds, 10 by default) and the computer to run it on.
  4. Choose Start warm-up.

The profile shows as launching, then running, while it works. When it closes and its data is saved the dialog says so. You can hide the dialog; the warm-up carries on.

A warm-up needs the profile closed, and the profile stays busy until it finishes: launches and edits wait for it.

From the CLI

terminal
veil warmup <profile-id> https://www.wikipedia.org/ https://www.bbc.com/ --dwell 15
# or keep the list in a file, one address per line
veil warmup <profile-id> --file sites.txt
SITE                        RESULT  TIME
https://www.wikipedia.org/  ok      17s
https://www.bbc.com/        ok      18s

Visited 2 of 2. The profile is closed and its data saved.

AI agents get the same through the warm_up_profile tool of the MCP server.

What it does on each site

  • Opens a new tab and loads the site, waiting up to 30 seconds for it.
  • Scrolls with real mouse-wheel input at irregular intervals until the dwell time is up.
  • Closes the tab. The profile's own tabs from its last session are left as they were.

A site that fails to load is skipped and reported; the warm-up only fails when none of the sites opened.

API

POST /v1/profiles/{id}/warmup with a dashboard session queues it, exactly like a dashboard launch:

{
  "deviceId": "<device-id>",
  "urls": ["https://www.wikipedia.org/", "https://www.bbc.com/"],
  "dwellSec": 15
}

It answers 202 with a commandId. Follow it at GET /v1/devices/commands/{commandId}: progress reads warming:<percent> while it visits, and status is done once the profile has closed and saved. Up to 50 sites per warm-up.

On the computer itself, the agent's loopback API runs one directly and answers when it is finished: POST http://127.0.0.1:38923/profiles/{id}/warmup with { "urls": [...], "dwellSec": 15 } and the agent's X-API-Key.

Was this page helpful?

On this page