Anthropic has announced the launch of a research preview for Claude for Chrome, a browser-based AI agent powered by its Claude models. The rollout begins with 1,000 subscribers on the company’s Max plan, priced between $100 and $200 per month, with a waitlist opening for additional users.
The Chrome extension allows users to interact with Claude in a side panel that maintains context across browsing sessions. With user permission, the agent can take limited actions in the browser to complete tasks. Anthropic emphasized that safeguards are in place, including restrictions on financial, adult, and pirated content, as well as permission checks for high-risk actions like publishing or sharing data.
The launch highlights the competitive race to integrate AI directly into browsers, with Perplexity’s Comet, Google’s Gemini integrations, and OpenAI’s rumored browser also shaping the landscape. Anthropic noted that browser agents raise new safety risks, but reported reducing prompt injection attack success rates from 23.6% to 11.2%. This research preview will serve as a testing ground to further strengthen defenses as Anthropic continues to advance agentic AI.
Featured image: Credit: Anthropic
