Let an agent act in your everyday browser
The extension surface: an agent works inside the browser you actually use, with your real logins, in new tabs, while you watch. When that's exactly right, and when it isn't.
Sometimes the thing you want done is in the browser you're using right now, signed into an account only you can reach, and watching it happen beats sending it off to a separate automated browser. That's what a browser-agent extension is for.
What makes it different
Every other surface in this series drives a separate browser. This one acts inside your everyday browser through an extension, using your real, live logins. It opens work in new tabs alongside what you're doing, and it asks you to grant permission on each site before it can act there. Claude-in-Chrome is one example of the category.
Best for
The task is in a site only you're logged into, in your daily browser. Or it's work you want to watch and steer as it goes, human in the loop. Or it's a quick one-off where standing up a separate automated browser isn't worth the setup.
Skip it for
Anything scheduled or unattended. This surface acts in your real browser and takes over tabs and focus, which is the opposite of what an unattended job should do (the safety rule from Part 1). Also skip it for high-volume or many-session work: it's one browser, yours, doing one thing at a time.
Cautions
It needs per-site permission, on purpose. Grant it deliberately, because it's acting as you, with your sessions.
Don't let it trigger the browser's built-in pop-up dialogs, the native alert, confirm, and prompt boxes. Those freeze the extension until you dismiss them by hand.
Give it a fresh tab for its work rather than the one you're reading.
This is the one surface where crossing into your daily browser is the point. Everywhere else in this series, keep automation out of the browser you live in.
Next up, Part 5: the opposite extreme, a browser with no login and nothing to remember.