Anthropic Launches Claude Browser Agent for Chrome

Anthropic Launches Claude Browser Agent for Chrome

Anthropic is pushing deeper into agentic AI with the launch of a research preview of Claude for Chrome — a browser-based AI agent designed to interact directly with the web. The company announced on Tuesday that the tool is rolling out to 1,000 Max plan subscribers (priced between $100 and $200 per month), with a waitlist now open for others.

What Claude for Chrome Can Do

By adding the extension, users can:

• Chat with Claude in a sidecar window that remembers browsing context
• Grant permissions for Claude to take actions in Chrome
• Offload certain browsing tasks to the AI, like filling forms or summarizing content

This marks Anthropic’s most ambitious move yet in bringing AI agents closer to everyday workflows.

The Growing AI Browser Race

Anthropic’s launch comes amid a fierce push among AI labs to dominate the browser space. Competitors are moving fast:

• Perplexity recently launched its AI browser Comet
• OpenAI is rumored to be close to debuting its own browser with agent features
• Google has integrated its Gemini AI models into Chrome

The timing is notable, as a federal antitrust case could force Google to sell Chrome, sparking speculation about potential buyers like Perplexity and OpenAI.

Safety First

With browser agents come new security challenges. Just last week, Brave flagged vulnerabilities in Comet’s agent to prompt-injection attacks. Anthropic says it has already built defenses, reducing attack success rates from 23.6% to 11.2%.

Additional safeguards include:

• Blocking Claude from accessing financial, adult, and pirated content sites
• Requesting explicit permission before high-risk actions (e.g., publishing, purchases, or sharing personal data)
• User settings to restrict access to specific websites

Why It Matters

Anthropic’s move highlights the evolution of agentic AI, where assistants go beyond chat to actively take actions for users. While early AI agents struggled with speed and reliability, today’s systems are becoming more capable at offloading routine tasks, though complex reasoning remains a challenge.

As Anthropic refines Claude for Chrome, the company hopes this research phase will help uncover and address safety risks before a broader rollout.