Let AI use a browser (four ways)
Some work only exists on a web page (a YouTube channel page, Amazon Seller Central). "Settings → Browser" says which browser the workstation uses. Leave it off and no role can open a page at all.
Three rules that never change:
- Each role can only open the sites its job description lists; everything else is refused: each creator channel its own platform, Amazon support only Seller Central, every other role none at all.
- Passwords are always typed by you in the browser — the AI never sees them.
- Screenshots go to the AI model you chose so it can see the page; they are not kept in the Agents Workshop records.
Which one
| Mode | In one line | When |
|---|---|---|
| No browser | The default. No role can open a web page | Work that doesn't need the web yet |
| Use the Chrome on my computer | The Chrome you're already signed in to: no second login, you clear the captchas; one task at a time | Normally |
| Use a separate Chrome | A clean browser for each task, signed in to nothing | The service isn't on your computer, or you don't want AI near your logins |
| The browser I already use | Your everyday browser (Chrome / Edge): anything you're signed in to is visible; the AI works in its own window and asks before touching a tab you have open | You need things only visible when signed in, without a second browser |
A separate Chrome never touches your everyday logins — clean, but you sign in to everything again. Your own browser can see what only a signed-in you can see; the AI works in its own window and asks before using a tab you already have open.
"Use the Chrome on my computer" and "The browser I already use" only work when the service runs on your own computer; on a shared company deployment they're greyed out with the reason (there 127.0.0.1 means the server's own machine).
Use the Chrome on my computer
- Choose "Use the Chrome on my computer".
- Click "Find it" to fill in the debugging address (default
http://127.0.0.1:9333), or type it and click "Test". - No debugging port open yet? Use "Open the work browser" in the desktop app tray — it opens a separate work Chrome and leaves your everyday one alone.
- Click "Save".
Use a separate Chrome
- Choose "Use a separate Chrome".
- Point at a Chrome / Chromium executable already installed, e.g.
/Applications/Google Chrome.app/Contents/MacOS/Google Chromeon macOS. We don't download a browser for you. - Click "Save".
The browser I already use (Tencent BrowserSkill)
Three extra steps, because two things can't be installed by us: the extension must be added by you in the browser, and bsk is a small local program.
- Install the extension in the browser you actually use (Chrome or Edge — either; nothing to configure): Chrome Web Store / Edge Add-ons.
- Install bsk (a small local program): click "Install". The extension needs it to talk to the workstation. Version and checksum are pinned: the download is verified first and nothing is installed if it doesn't match.
- Check it: whether the local service is running and the extension is connected; if not, each item tells you how to fix it. "All set" means roles can start working.
- Click "Save".