Anthropic releases Claude Duet personal agent built on Opus 5.5

Anthropic has released Claude Duet, a personal agent product designed for professional and small business users. Currently available as a research preview via waitlist, the tool is built on the Opus 5.5 model. Duet differentiates itself through proactivity, sending morning briefs on connected applications and alerting users only when their attention is required. The system supports a range of tasks, from scheduling to building dashboards and web apps, and features a virtual machine, persistent memory, and a dedicated safety classifier. This classifier requires explicit permission for major actions such as payments or file deletion.

Anthropic warned that the early version may misread threads or miscalibrate urgency, noting it is less capable at personal errands like booking tickets. Community reaction highlighted the underlying Opus 5.5 model as the key differentiator. The release positions Anthropic in a crowded market alongside Meta's Muse, OpenAI's Dots, and SpaceX's Bots. Observers noted that competition is shifting from model intelligence to owning the persistent user relationship. Commentators view the launch as a strategic move by Anthropic to secure user relationships before a potential IPO, signaling that peak intelligence for most workflows has been reached and the focus is now on product integration. One observer noted that pretty much every major AI company now has a personal agent product except Google.