Anthropic has taken another step in integrating its artificial intelligence with the browser. The company has launched a Chrome extension that allows Claude to see the page the user has open and perform actions on it, such as clicking, typing text, or navigating between sites. This new feature, known as Claude Cowork, transforms the assistant into an agent capable of executing tasks directly in the browser, without the need to switch tabs or applications.
The extension is now available in the Chrome Web Store for Max and Team plan subscribers, while Pro plan users will have to wait a few more weeks. On Enterprise accounts, the feature is disabled by default and must be enabled by administrators. Additionally, working with local files still requires using the Claude desktop application.
Claude Cowork is Anthropic's tool designed to perform multi-step tasks . With this extension, the assistant can access the content of open tabs, extract information, fill out forms, and complete processes that previously required manual intervention. For example, it can be asked to review several open invoices, extract the amounts and dates, and generate a spreadsheet with the data. All of this is displayed directly in the Chrome sidebar.
One of the most notable new features is that conversations started in the browser are saved in Claude's history and can be continued from other platforms, such as the desktop app, the web, or mobile. This means that a task started in Chrome can be resumed later on another device without losing context. Skills and connectors configured in the account are also accessible from the browser.
Availability and requirements
The extension can be installed from the Chrome Web Store, but it doesn't work on other Chromium-based browsers or mobile devices. According to the store listing, it's a beta version . Max and Team plans already have access, and Anthropic has announced that Pro users will be able to use it in the coming weeks. For businesses, administrators can enable the feature and restrict its use to authorized domains.
To get started, simply install the extension , log in with your Claude account, and open the sidebar. From there, you can write clear instructions for the agent to work with the tabs open. However, for tasks that require local files or integrations with other applications, you'll still need the desktop application.
Safety and risks
Allowing an AI agent to operate within the browser carries risks, particularly prompt injection , where hidden instructions on a webpage can attempt to mislead the assistant. Anthropic has implemented safeguards, such as an independent check that reviews actions when automatic approval mode is enabled. Furthermore, Claude still requires confirmation for sensitive actions like making purchases or sharing personal information with the AI.
The company acknowledges that these measures reduce the risk but do not eliminate it entirely. Therefore, it recommends starting with trusted sites, reviewing granted permissions, and maintaining control over the most sensitive tasks. In enterprise environments, lists of allowed and blocked domains can be established to limit the scope of the extension.
Despite the risks, Claude's integration into Chrome represents a significant step forward in web task automation. Users can delegate repetitive processes, such as collecting data from multiple pages or filling out forms, with the confidence that the system has safeguards in place. The key is to use the tool with caution and always review the results.
Claude's new Chrome extension transforms the assistant into an agent capable of interacting with the browser and performing tasks autonomously. With cross-device synchronization, availability on Max and Team plans, and implemented security measures, Anthropic aims to deliver a smoother and more productive experience. While some challenges remain, this integration marks a milestone in the evolution of AI assistants.
