Anthropic Launches “Claude Duet,” a Proactive AI Assistant Agent

 Anthropic has launched Claude Duet, a research preview of a proactive AI assistant agent that monitors connected apps, sends morning briefings, and completes tasks in the background. It runs on Claude Opus 5.5 and is available to a limited number of Pro and Max subscribers in the U.S. through a waitlist, positioning Anthropic against OpenAI’s Dots and Meta

Anthropic has entered the AI agent race, and it did so with a product that behaves less like a chatbot and more like a digital secretary who starts working before you ask. The company announced Claude Duet on Wednesday, opening a research preview for U.S. subscribers on its Pro and Max plans through a waitlist.

The central idea behind Duet is simple and slightly unsettling in the best way: you should not have to initiate every interaction. Anthropic built the agent to monitor connected applications, surface priorities, and carry out assignments in the background, even after you close your laptop or turn off your phone. Each morning, Duet generates a short briefing with action items drawn from your email, calendar, and business software, then checks those accounts every few hours and reaches out only when something needs your attention.

That shift from answering questions to running parts of your working life is the story here. Claude Duet is Anthropic’s most direct attempt to move its chatbot from a responsive tool into an agent that initiates, prioritizes, and acts. It runs on Claude Opus 5.5 and operates in its own virtual machine with a browser, so it can keep working remotely without needing your device to stay awake. It carries over the memory, skills, and app connections from your existing Claude account, and it supports more than 6,000 integrations, including Gmail, Slack, QuickBooks, Stripe, HubSpot, and Shopify.

The range of tasks is broad. Duet can schedule meetings, investigate unpaid invoices, draft customer follow-ups, build financial forecasts, create dashboards, and even prototype web apps. But Anthropic is careful to frame Duet as a work tool first. Unlike OpenAI and Meta, which have pitched their agents as assistants for everyday tasks like booking reservations, Anthropic says Claude Duet was built for work and is less capable at personal errands.

Safety is the other half of the pitch. Duet asks for approval before sending messages, making payments, signing documents, or deleting files, and it blocks actions like trading securities or transferring money between accounts altogether. Anthropic says the agent never directly sees your passwords or sign-in tokens, and a new classifier and safety system prompt guard against leaks of sensitive information, including credit card numbers and government IDs. The company has also subjected Duet to red-team testing and an external penetration test.

There are limits, and Anthropic is upfront about them. Duet can misjudge how urgent a message is, personal errands like booking tickets are still a work in progress, and users cannot yet watch the agent’s browser as it works. Early reviewer Dan Shipper of Every described Duet as a professional tool but found its email drafts weaker than a competing agent, and noted that voice features are limited to transcription for now.

The competitive framing matters. OpenAI introduced its Dots personal agent in September, and Meta launched Muse earlier that month. Anthropic is later to the party, but it is entering with a distinct position: a work-first agent bundled into the existing Pro plan at no extra cost, with background execution that does not require a persistent session. Whether that is enough to set Duet apart will depend on how well its judgment holds up as the waitlist expands.

For now, Claude Duet is a research preview, not a general release. Access opens in waves, and Anthropic says features can change during the preview. But the direction is clear. The chatbot era was about answers. The agent era, as Anthropic is now betting, is about initiative.

’s Muse.