Skip to content

Anthropic

Claude’s Chrome extension now shares one session across your devices

3 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

Video: threads.com

Anthropic on August 12 turned the Claude in Chrome side panel into a full Claude Cowork client. The panel now runs the same session as the desktop, web and mobile apps, and that session is tied to your account instead of the machine you started on. The change was announced in a post from Anthropic on Threads. Max and Team subscribers have it now, and Pro gets it “in the coming weeks.”

Conversations you start in the browser now go into your regular Claude history. You can begin a task in a Chrome tab and finish it in the mobile app, or start on the desktop and continue in the sidebar. The skills and connectors you have already set up also work in the browser panel.

Until this update, the extension was “more or less its own thing,” almost completely separate from the other Claude apps, and that was one of the main complaints in its Chrome Web Store reviews, The New Stack reported. The outlet also notes that Cowork reached web and mobile only a few weeks earlier, after living only in the desktop app, and that this work was likely needed before it could run in the extension.

The extension’s advantage over the other Claude apps is still that it can operate the browser and see the page you are on. Anthropic, as quoted by The New Stack, frames it this way: “Many of the tools you use every day connect directly to Claude, but others don’t, such as internal dashboards, legacy systems, and vendor portals. With Claude in Chrome, Claude can work in these apps through the browser.” With connectors in the same panel, one session can now apparently cover both cases: connected tools through their integrations, and everything else by clicking through the page.

Pro subscribers have had the extension since December 18, 2025, when Anthropic opened it to Pro, Team and Enterprise. What reaches Pro in the coming weeks is the Cowork version. The same December rollout gave Team and Enterprise admins org-wide on/off switches plus site allowlists and blocklists. The New Stack says that on Enterprise plans the new experience is off by default, and admins can turn it on and limit it to specific domains. The outlet adds that the extension still runs only in Chrome itself: not in other Chromium-based browsers, and not on mobile.

Anthropic’s post warns that browser agents “can be tricked by instructions hidden in a page.” This attack is called prompt injection: a site, email or document carries text meant to steer the agent away from what the user asked. The company says it builds defenses and still recommends some habits of your own, which it lists in a support article on using Claude in Chrome safely.

Anthropic has measured this risk before. The original pilot went to 1,000 Max users, and in red-teaming across 123 test cases covering 29 attack scenarios, browser use without safety mitigations let attacks succeed 23.6% of the time; the defenses added then cut that to 11.2%. The extension moved to a beta for all Max subscribers on November 24, 2025, and Anthropic has since reported a 1% attack success rate against its internal Best-of-N attacker, which got 100 attempts per environment.

For this update, The New Stack describes Anthropic as extending Claude’s auto mode to the browser agent, with an agent automatically checking for anything out of the ordinary so the user does not have to approve every action. Claude still asks before making purchases or sharing personal data. The outlet quotes Anthropic’s own caveat: “While these measures meaningfully reduce the risk, they cannot eliminate it.”

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.