Anthropic starts limited Claude for Chrome pilot with safety controls

Anthropic has begun a limited research preview of Claude for Chrome, a browser-based AI agent designed to operate inside the user’s web window. The pilot targets 1,000 subscribers on the Claude Max plan, with access granted via a Chrome extension that places Claude in a sidecar chat window and allows the agent to monitor browser activity and perform actions on the user’s behalf. A waitlist is open for additional users as the company collects feedback on safety, reliability, and practicality.

The Claude for Chrome extension lets the AI see webpages, click buttons, and fill out forms, enabling it to carry out multi-step tasks across sites. Demonstrations show Claude handling tasks such as scanning Gmail for unresponded messages and drafting follow-ups, filtering listings on real estate sites, and even adding items to a shopping cart. In practice, users authorize Claude to access specific sites and grant permission before executing high-risk actions, such as sending messages, publishing content, or making purchases. The feature is positioned as a natural extension of Claude’s existing capabilities to interact with calendars, documents, and other apps already connected to the assistant.