Anthropic Enters the AI Agent Market with Claude Duet

The agent operates by aggregating daily briefings from connected applications, alerting users only when specific tasks require human intervention. Beyond simple scheduling, Duet handles complex administrative functions including dashboard construction, financial forecasting, and web application prototyping. It leverages a virtual machine environment equipped with a browser to execute these tasks, maintaining a human-in-the-loop requirement for sensitive actions like financial transactions.

Security remains a focal point for the release. Anthropic has implemented a dedicated classifier and safety system designed to block requests involving high-risk data, such as credit card numbers. Despite these protections, the company categorizes the current build as an early version, acknowledging potential limitations in prioritizing urgent communications or executing personal errands like travel bookings. Users can now access the research preview to test the agent's ability to navigate integrated professional tools.