Think about how much of your job actually happens inside a browser tab.
Email, calendars, docs, the expense portal nobody likes.
For most people who work on a computer, the browser is the job.
Until recently, AI could only talk about that work. You'd describe what you wanted, copy something out, paste it somewhere else, and do the clicking yourself.
Browser use changes that. Now Claude can sit in your browser, see what you see, and do the clicking, typing, and form-filling for you.
Today we're starting simple: what Claude in Chrome is, how to turn it on, and the first task to try so you can feel the difference for yourself 👇
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
What "browser use" actually means:
Regular Claude lives in a chat window. It can answer questions about a website, but it can't go there.
Claude in Chrome is different. It's a free browser extension that adds a Claude panel to the side of Chrome. That panel sees the page you're on and can take action on it (navigate to a site, read what's there, click buttons, fill out forms, work across several tabs at once).
You ask in plain English; it works the page like a person would.
A few things worth knowing before you install:
It runs on Chrome only (not Safari, Edge, or other browsers) and on desktop, not mobile.
It's in beta for all paid Claude plans (Pro, Max, Team, and Enterprise)
On Pro, it runs on Haiku, Anthropic's fast model. On Max, Team, and Enterprise, you can pick a more capable model for trickier tasks.
How to turn it on 👇
Open Chrome and go to the Chrome Web Store.
Search for Claude and click Add to Chrome on Anthropic's extension.
Sign in with your Claude account.
Pin it: click the puzzle-piece icon in the toolbar, then the pin next to Claude.
Grant the permissions it asks for (this is what lets it read pages and click for you).
From there, a Claude icon lands in your toolbar.
Click it and a panel opens on the right that stays put while you browse (that panel is where you'll type your requests).
Your first task: let it read the screen.
The fastest way to get browser use is to point it at something already on your screen.
Open a long article, a dense pricing page, or a report you've been avoiding. Open the Claude panel and type:
Summarize this page for me in five bullets, then tell me the one thing I should actually do about it.
It reads the page you're looking at and answers straight from it. Small, but it's the whole idea in miniature: Claude is working with the actual page, the way you see it.
Working in multiple tabs
This is the part that stops feeling like a chatbot.
Say you've got five tabs open for a decision (three vendor pages, a review site, and a competitor).
Normally you'd flip between them and hold it all in your head.
Instead, drag those tabs into Claude's tab group (it color-codes its own tabs so you always know which are yours).
From there, you can ask something like:
Compare these five tabs on price, what's included, and the catch in the fine print. Put it in a table and tell me which one you'd pick for a small team.
Claude reads across all of them and hands you one answer.
The thing that used to eat twenty minutes of tab-flipping takes one 10-second request.
Claude already knows its way around the tools most people live in (Gmail, Google Calendar, Google Docs, Slack, GitHub) so requests like "find the thread about the Q3 budget and draft a reply" work without you spelling out every step.
And now, let’s automate 💪
Doing a task once is handy. The payoff is not having to do it again.
Two features turn a one-off into a standing routine:
Shortcuts. When a prompt works well, save it. Type / in the panel to pull up your saved shortcuts and run the same thing instantly. Your "Monday catch-up" or "summarize this and draft a reply" becomes one click.
Scheduling. You can set a shortcut to run on its own (every morning, every Monday, monthly). Claude does the task on schedule and pings you when it's done.
For anyone whose week is full of small, repeating browser chores, it's where the hours come back.
READ THIS BEFORE YOU LET LOOSE
Handing an AI the keys to your browser is powerful, and it deserves a few guardrails. Anthropic builds in several, and you should use them:
You control which sites it can touch. Grant or revoke access per site in Settings.
It asks before risky moves. Publishing, buying, or sharing personal data triggers a confirmation step. There's an "Ask before acting" mode that has Claude show you its plan first, then carry it out within those bounds.
Some sites are off-limits by design. Anthropic blocks whole categories like financial services and adult content.
The risk to understand is prompt injection: a malicious page or email can hide instructions that try to hijack an AI reading it (think phishing, but aimed at your AI assistant instead of you).
It's a real, active area. A flaw nicknamed ClaudeBleed surfaced this spring, and Anthropic shipped a fix.
Start on sites you trust, keep an eye on what's visible to Claude, and steer it away from anything financial, legal, or medical while you're learning what it does well.
Paste this into the Claude in Chrome panel on a page or set of tabs you're comfortable sharing.
Fill in the bracketed section and let it go.
I want help with what's on my screen right now. The task:
[Describe what you want — e.g., "compare these open tabs," "summarize this report and draft a reply," "pull the key dates out of this page."]
First, tell me back in one line what you think I'm asking, so we're aligned.
Then do the task using only what's on the pages I've shared with you.
If a step would publish, send, buy, or share anything, stop and check with me before you do it.
When you're done, give me the result plus one suggested next step.Run it once. If it works, save it as a shortcut (and you've built your first piece of browser automation 🙌)
Your roundup of the latest model releases and updates from the biggest AI labs.
Anthropic files to go public. The company behind Claude filed confidential paperwork with the SEC to launch an IPO: the first step toward selling shares on the stock market. Recent fundraising valued it around $965 billion, ahead of OpenAI, and CNBC reports its revenue run rate (annualized sales pace) has jumped to roughly $47 billion, up from about $10 billion a year ago. The company is set to become one of the most valuable public companies in tech. (CNBC)
Microsoft and Google start building their own AI coders. Both giants are rolling out homegrown AI coding models to chase Anthropic and OpenAI, and Microsoft is expected to show its in-house models at its Build conference this week. Anthropic's Claude Code pulled ahead of Copilot with developers, and Microsoft wants the lead back. For everyone else, it means more capable AI assistants competing for your everyday work. (CNBC)
Apple's revamped, Gemini-powered Siri is expected next week. Previews ahead of Apple's June 8 developer keynote point to a rebuilt Siri running partly on Google's Gemini, with the ability to understand what's on your screen, take actions inside apps, and hand questions off to assistants like Claude or Gemini. If it lands, it puts a genuinely capable AI assistant in front of roughly two billion devices. It could be the moment that "AI that does things for you" goes mainstream. (Tom's Guide)
Advertise with Build with AI
Get in front of an audience of professionals using AI day-to-day: founders, engineers, operators, and product builders.
Interested in advertising? Respond to this email for rates and details.
Until next time,
William Ryan
Editor-in-Chief @ Build with AI
PS: Follow me on X for daily updates and AI workflows.