Now loading...
The integration of Claude into the Chrome side panel has evolved into a full-fledged Claude Cowork session, enabling users to have their conversations saved in a history log. This new functionality allows users to begin a task in a tab and seamlessly continue it across Claude’s desktop, web, and mobile applications. Currently available for users on Max and Team plans, it will soon be accessible to Pro users in the following weeks.
The Claude in Chrome extension provides the AI with the ability to interact directly with the web pages users visit. It can perform actions such as clicking on links, typing inputs, navigating through different sections, and filling out forms using existing login information.
Although many everyday tools connect directly with Claude, some specific platforms—such as internal dashboards and vendor portals—did not have this capability. With the latest extension, Claude can now engage with these applications through the browser interface.
Previously, sessions conducted in the side panel operated independently from those in Claude’s other applications, thus lacking continuity in context and dialogue. The enhancement means that now the side panel utilizes the same Claude Cowork session that is accessible across desktop, web, and mobile platforms, facilitating more extensive workflows that span multiple steps. As sessions are tied to individual accounts, users can simply switch from working in a browser to another device later.
For instance, users creating a budget spreadsheet that requires invoice data from various vendor portals can instruct Claude in Chrome to gather amounts and dates. The AI will open the relevant tabs, extract data from each invoice, and compile the spreadsheet. Users can continue the session on the desktop app, where they can incorporate files from their computer or review the previous month’s budget to identify changes, thus ensuring that the context remains intact throughout their work.
It is important to note that using Claude in Chrome carries inherent risks associated with AI operations in browser environments, particularly the threat of prompt injection. Malicious individuals can embed harmful instructions within web content that are not immediately visible, leading Claude to perform unintended actions.
To mitigate these risks, enhancements have been made to how Claude handles its tasks following the initial pilot. With the ‘automatically approve’ function, it can execute tasks without repeated permission requests. However, before carrying out significant actions—such as submitting forms or downloading files—a verification system ensures that the actions align with user requests, thus reducing unnecessary interruptions while keeping safety measures in place.
Claude will still seek approval before executing high-stakes actions such as making purchases or sharing sensitive information. While these precautions substantially lessen risks, they’re not infallible. The dynamic nature of prompt injection keeps developers vigilant, consistently looking for new threats and integrating learned defenses into future model updates. Users are advised to begin operations on trusted websites, and more best practices can be found in the safety guide.
To start using Claude in Chrome, users should download it from the Chrome Web Store, sign in, and access the side panel. The new panel is currently available for Max and Team plan users and will soon be more widely accessible to Pro users. For organizations on Enterprise plans, Claude in Chrome is disabled by default; administrators can enable it and restrict it to approved domains. Additional setup guidance for administrators can be found in the admin setup guide.
It remains essential to use the Claude desktop application for file management or interaction with other software, as Claude in Chrome is not yet compatible with other Chromium-based browsers or mobile devices.
