Summary
Anthropic has introduced Claude Duet, a research-preview AI agent that works across connected apps, prepares daily briefs and handles tasks in the background. Access is currently listed for Claude Pro and Max subscribers in the US.
Anthropic has introduced Claude Duet, a proactive AI agent in research preview that can monitor connected apps, prepare a daily brief and carry out tasks after receiving instructions. The company says the service is powered by Opus 5.5 and is designed to keep working in the background, then return completed work or ask for a decision when needed.
Unlike a chat assistant that mainly responds when a person opens a conversation, Duet is intended to surface items that may need attention and take on multi-step work. Anthropic says users can message it by text or voice, with the conversation, memory, skills and connected apps from their Claude account available in one thread.
How Claude Duet is designed to work
Duet runs tasks on a computer in the cloud and uses apps connected by the user. Anthropic describes uses including drafting email replies, preparing payroll for review, finding unpaid invoices, and assembling documents, presentations or dashboards. It says the product works with more than 6,000 connectors and plugins.
The agent can also provide a morning briefing based on information in connected inboxes and calendars, and follow up on tasks it has been given. Anthropic’s examples show it preparing work for review and asking before actions such as sending a message or making a payment. The company says approvals, logins and safety checks for sensitive operations run in the cloud computer, and that users can see and steer what Duet is doing.
The service is offered as a standalone app for iOS and Android, and can be opened on the web at claude.ai. Anthropic lists it as available through a waitlist for Claude Pro and Max subscribers in the US; Team and Enterprise plans are not included yet. The company has not specified a separate price for Duet. During the research preview, it shares the subscriber’s Claude usage limits, and background checks count toward those limits. Larger tasks may use more model turns and background work.
Research-preview limits
Anthropic describes Duet as an early research preview and identifies specific reliability risks: it may misread a conversation thread, judge urgency incorrectly or miss something that arrives between check-ins. Those caveats matter because the product is designed to act on information across connected services and decide when to alert its user. Its approval steps provide a review point for sensitive actions, while the company’s stated limitations mean the briefings and prioritisation still require oversight.