Anthropic introduced Claude Duet, an agent for professionals and small businesses that monitors connected apps and handles tasks with user approval.
Anthropic introduced Claude Duet on October 21, 2026, as a research preview aimed at professionals and small business owners. The agent is available through Claude on the web and as a standalone app for iOS and Android. Access is limited to people in the United States with a Pro or Max subscription, and interested users must join a waitlist.
Built around Claude Opus 5.5, Duet can connect to apps and use information from Claude’s memory and skills. Anthropic describes a workflow in which the agent prepares a morning summary from connected services, keeps watch during the day, and sends a message when something appears to need attention. Users can also assign work ranging from scheduling to creating dashboards, forecasts, prototypes, and web applications.
The agent can use a browser in a virtual machine to carry out tasks. For consequential actions—including sending a payment or deleting material—it asks the user to approve the step first. Anthropic has also added a dedicated classifier and safety prompt intended to reject requests to share highly sensitive information, such as credit card numbers.
The preview comes with stated limitations. Anthropic says Duet is an early version that may misunderstand message threads or judge urgency incorrectly, and that it is less capable at personal errands such as booking tickets. The product therefore pairs ongoing app monitoring and task execution with approval checks, while leaving users to evaluate how reliably it handles their work.
The launch places Duet in a developing field of assistant-style agents, with The Verge noting Meta’s Muse and OpenAI’s Dots as competitors. Anthropic’s initial rollout is a waitlisted US preview for eligible subscribers, rather than general availability.