Claude for Chrome goes live for early subscribers

Anthropic tests a browser based Claude agent that can view and act inside a user’s Chrome session.

Anthropic launches Claude in Chrome browser

Anthropic is rolling out a research preview of Claude for Chrome, a browser based AI agent that can view and act within a user’s Chrome session. The feature, Claude for Chrome, is available to about 1,000 subscribers on Anthropic’s Max plan, priced between 100 and 200 dollars per month. The agent works through a Chrome extension and runs in a sidecar window that preserves context as the user browses. Users can grant Claude permission to take actions in the browser to complete tasks.

The move places Anthropic in a crowded field as AI labs race to embed agents in browsers. Perplexity has launched Comet, and OpenAI and Google are pursuing browser integrations. The case around Google Chrome has added a political edge to the tech race, with antitrust scrutiny looming. Anthropic warns that browser access creates new safety risks; Brave security researchers flagged a vulnerability in Comet's browser agent, though Perplexity says it has fixed the issue. Anthropic notes that it has deployed defenses that reduced the success rate of prompt injection attacks from 23.6% to 11.2%.

This launch signals a shift toward AI helpers that live inside the browser, turning everyday browsing into possible automation. That convenience comes with questions about data handling, user consent, and how much control people should grant an agent that can interact with pages. The real test will be whether safety measures keep pace with capability and whether users trust the system enough to let it perform tasks.

Competitors are racing to embed similar agents, raising regulatory and privacy concerns as browser ecosystems come under scrutiny. The blend of fast growth and high risk means users and investors will watch closely for safety, transparency, and reliable performance as the space evolves.

Highlights

The path forward will depend on how well safety and user trust can keep up with rapid capability growth.