The setup

Claude for Chrome

Anthropic's official extension puts Claude inside your browser, logged in as you. How to set it up, and what it's actually worth using for.

Start here: install it

It's in beta on any paid Claude plan. Two minutes, works in Google Chrome (not Edge, Brave, Arc, or mobile).

1

Open the Chrome Web Store listing

It's the official one, published by Anthropic. The link is just below.

2

Click “Add to Chrome”, then pin it

Pin the extension so the Claude side panel is one click away whenever you need it.

3

Sign in with your Claude account

The same login you use for Claude. Now Claude can see and act on whatever tab you're on.

Get it from the Chrome Web Store

Your first task

Open the side panel and paste this in. It's a safe one to start with, read only, it just looks and reports back rather than sending anything.

Paste into the Claude side panel
Here's a first task to try you out. Go to my LinkedIn messages, look through my recent conversations, and give me a list of the people I haven't replied to yet, with a one line summary of what each of them last said. Don't send anything or reply to anyone. Just show me the list so I can decide what to do.

What it's actually worth using for

The point of Claude for Chrome is the stuff behind a login that a normal integration can't reach. Because it's working in your own signed in browser, it gets to everything you get to.

Forms and portals

The quarterly renewal, the expense claim, the supplier portal nobody remembers the password for. It fills them in for you.

Staying on top of messages

Point it at LinkedIn or your inbox and have it list who you've left on read and what they last asked, so nothing important slips.

Pulling things together

Grab the numbers off a dashboard, compile a list across a few open tabs, organise files in a drive, no copy and paste.

On a schedule

Set a task to run daily or weekly from the clock icon in the panel. One honest catch: it only runs while Chrome is open, it won't wake your laptop for you.

The honest bit, read this before you let it loose

  • -It's still an AI and it's still figuring things out. Pick the jobs you'd be comfortable handing over, and use the “ask before acting” setting so it shows you the plan before it does anything. Here's how to decide which tasks to trust it with.
  • -Anthropic themselves say don't use it for financial transactions, password management, or anything with sensitive personal data. Start on trusted sites and grant permissions carefully.
  • -There's a real risk called prompt injection, where a dodgy web page hides instructions to try and trick the agent into doing something you didn't ask for. It's why you keep it on sites you trust and watch what it does the first few times.
  • -If there's a proper connector or integration for something, use that instead. It's cleaner and more reliable. Claude for Chrome is for everything behind a login that hasn't got one.