Anthropic has expanded Claude’s capabilities in Chrome, enabling seamless cross-device conversations and more direct page interactions, while introducing new security measures amid concerns over vulnerabilities.
Anthropic has expanded Claude in Chrome so it no longer behaves as a separate browser-only tool. The extension now keeps a shared Claude Cowork session across desktop, web and mobile, allowing conversations, skills and connectors to follow the user between devices. That is a notable change for people who begin a task in the browser and want to finish it later in the desktop app without starting again.
The browser version can also act more directly on a page. According to Digital Trends, Claude can read the site a user is on and carry out actions such as clicking links, entering text, completing forms and moving through pages while using existing logins. That makes it more useful for work on internal systems, older business software and vendor portals that do not integrate cleanly with AI tools. The publication said the feature is now live for Max and Team plans, with Pro access due in the coming weeks.
Anthropic is not the only company pushing Claude deeper into the browser. Reporting from 9to5Mac said the company has also introduced an in-app browser for Claude Code on desktop, aimed at letting developers consult documentation and interact with web pages without leaving the coding environment. TechCrunch previously reported that Claude for Chrome began as a research preview for a limited group of Max subscribers, showing that Anthropic has been steadily broadening the product from a narrow trial into a more integrated assistant.
The wider rollout comes with sharper security controls, but also clearer warnings. Anthropic has added a review step for consequential actions so that requests are checked against the user’s intent before they are carried out, and Claude still asks for permission before purchases or sharing personal information. Even so, TechRadar reported that security researchers at Manifold Security found serious flaws in version 1.0.80 of the Claude for Chrome extension, including a way for malicious extensions to trigger Claude workflows and bypass permission prompts. That tension reflects the central problem of browser automation: the more capable the assistant becomes, the more damage it can do if it is misled by hidden instructions on a page.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
