Anthropic Opens Claude Chrome Extension to All Paid Users With Powerful Automation Features

Anthropic Rolls Out Its Chatbot Claude as a Chrome Extension

OpenAI rival Anthropic has now rolled out its chatbot Claude as a Chrome extension for all paying subscribers in beta. The AI firm says the new extension will be able to navigate websites on your behalf, fill in forms, and complete scheduled tasks, accessing multiple tabs in the process. You'll need to pay a minimum of $20 per month for a Pro subscription to access the Claude for Chrome extension, an $80 downgrade from when Anthropic started beta testing the feature exclusively for $200-per-month Max users in November. If you're not a Chrome fan, the tool is also available on alternative browsers like Brave and Opera, which use the same Chromium engine for their underlying infrastructure. Firefox, Safari, and Microsoft Edge are not officially supported, however. Anthropic acknowledged the "safety and security challenges" that come with incorporating chatbots directly into your browser experience, including the risks of what is known as prompt injection attacks. This is where hackers try to input deceptive and malicious text into a large language model (LLM) like Claude. The company detailed some of the safety features it included to protect against these types of attacks, for example, the ability to control site-level permissions, where users can grant or revoke Claude's access to specific websites at any time via the Settings tab. In addition, the extension will also ask users before carrying out "high-risk actions" like publishing content, purchasing things online, or sharing personal data. It will also automatically block itself when accessing high-risk websites, for example, financial services websites such as your bank, pornography, or pirated content such as torrent databases. Anthropic certainly isn't the only firm incorporating what's called "agentic AI," where AI does things on your behalf (as opposed to just answering questions), directly into browsers. OpenAI rolled out Atlas in October of this year, offering broadly comparable features to the new extension, though it also requires a Premium subscription to use. Meanwhile, the ChatGPT maker's CEO Sam Altman has been extolling "agentic AI" as the future, as have figures like Nvidia CEO Jensen Huang. And if you don't feel like forking out a monthly subscription for your AI browser, Perplexity Comet started rolling out for free users in October.

Claude's Chrome plugin is now available to all paid users

The Claude Chrome plugin allows for easy access to Anthropic's AI regardless of where you are on the web, but its real draw is how it lets Claude navigate and use websites on your behalf. Anthropic says that Claude can fill out forms, manage your calendar and email and complete multi-step workflows based on a prompt. The latest version of the plugin also features integration with Claude Code, Anthropic's AI coding tool, and allows users to record a workflow and "teach" Claude how to do what they want it to do. Before agents were the buzzword du jour, "computer use," the ability for AI models to understand and interact with computer interfaces, was a major focus at Anthropic and other AI companies. Now computer use is just one tool in the larger tool bag for agents, but that understanding of what digital buttons to click and how to click them is what makes Claude's Chrome plugin possible. OpenAI and Perplexity offer similar agentic capabilities in their respective ChatGPT Atlas and Comet browsers. At this point the only AI company not fully setting its AI models loose on a browser is Google. You can access Gemini in Google Chrome and ask questions about a webpage, but Google hasn't yet let its AI model navigate or use the web on a user's behalf. Those features, first demoed in Project Mariner, are presumably on the way.

I tried the new Claude in Chrome extension, and it delivered convenience with a side of digital paranoia

Anthropic has opened access to the Google Chrome browser extension version of its Claude AI assistant to Claude Pro, Team, and Enterprise subscribers, following several months of testing with subscribers to the highest tier, Max. Claude in Chrome actively works to navigate the web, complete tasks across tabs and even perform coding duties, not just summarize websites.. The idea is to place Claude in areas it previously couldn't reach. Instead of asking an AI to explain what you're seeing, Claude can now see it too and interpret and interact with live webpages. I gave the new version of Claude a spin. I activated the extension and assented to giving the AI access to basically everything I do online. That gave me a moment of trepidation, but the broad permission to observe and interact with my browsing activity pretty well defines the Claude in Chrome experience. Once authorized, a small Claude icon appeared next to my Chrome address bar. I clicked it and saw the familiar chat interface slide into view. There are a lot of tasks you can assign to Claude in Chrome. I had it set up a meeting by giving a brief description of when, where, and with whom I wanted to meet. The AI scanned my Google Calendar and proposed open time slots within a draft email to my guests. I also took a page from Anthropic's demos to see how well the AI did at organizing my incredibly scattered Google Drive. After giving it access to the platform, I asked it to go through and put the hundreds of documents and spreadsheets into some order, arranging them in folders with related documents. The AI did exactly that, setting aside any it wasn't sure of for me to assign appropriately. I went from close to 900 loose documents and spreadsheets to six top folders with several subfolders, and nearly 50 duplicates were tidied away. Arguably, the most intriguing and powerful feature offered by Claude in Chrome is recording a workflow to teach the AI how to do something on your behalf. You click the record button, go about your business as usual, and Claude watches and remembers. The tabs and menus you open, the forms you fill out, and the ultimate goals are all absorbed and become a repeatable pattern that Claude. When you stop recording, Claude doesn't just memorize clicks; it understands the sequence as a task it can repeat when asked. I recorded myself running a series of checks across various reward programs I belong to and logging the results in a spreadsheet. When I asked Claude in Chrome to "run the reward points check workflow," it followed my instructions. Claude handled it quite well, with only a couple of errors in passwords slowing it down. The AI even offered to make it a monthly routine that I won't even have to check up on. And despite the many tabs required for the project, Claude kept them all in its little workspace, running everything in parallel without my having to supervise even the password submissions. That undefined boundary between assistance and oversight is where I started to feel like I might have shared too much with the AI. Automating interactions, especially those involving passwords, means exposing personal data to Claude. Every new permission widens the door for Claude to walk through. And since I'm not watching the whole time, I might not even know I've been logged into an account. There's nothing sneaky about it. When you start an automation workflow, Claude asks for permissions. That said, the way it stitches itself into your web experience leaves you wondering how much of your digital life you've opened up to Claude's inspection. That's the kind of capability that makes Claude in Chrome stand out as more useful than lighter AI extensions limited to visible text. Comparing Claude in Chrome with other AI browser extensions makes that distinction clearer. Tools attached to models like ChatGPT often limit themselves to helping reshape text or summarize selected articles. Claude is designed to infer context and act upon it. On the other hand, Claude's deep dives mean it's acting as your representative, and you're not even nearby. I might have tried Claude in Chrome even if it weren't for an article, but it did bring home to me that as "AI agents" become more ubiquitous and more powerful, the question of trusting AI with your data will matter for how it is deployed on the internet in real time as much as how it is absorbed by the models powering AI assistants. Anthropic acknowledges this and even notes that people shouldn't use the automated version of Claude for tasks like banking. Still, with the right privacy protections in place, I can see using Claude to handle all the tedious chores I have online, though I'll still want to check up on it every so often.