Claude Duet hands-on: Anthropic's agent nags, but catches missed deadlines

Claude Duet, Anthropic's agent for people who don't write code, spent a week reading a reporter's work email, calendar, Google Docs and Slack. Allison Johnson of The Verge found it useful at catching what she had missed and too eager to rush her.

“I like it as a backstop for my own organizational systems, not as a replacement for them.”

— Allison Johnson, reviewer at The Verge, Claude Duet hands-on

Quote source: The Verge, Allison Johnson's hands-on review, 13 November 2026.

What Duet is

Anthropic introduced Duet on 21 October as a research preview for professionals and small business owners. It runs as a separate app on iOS, Android and claude.ai, with a waitlist for US subscribers on the Pro, Max and Ultra plans, and it shares their usage limits.

Every morning Duet sends a short brief and then suggests tasks as cards you swipe to accept or dismiss. It checks connected accounts every few hours and does the work in its own virtual machine with a browser, so your phone can stay off.

Anthropic set hard limits. An AI agent like this one asks before it sends a message, pays, signs a document or deletes anything. It will not trade securities or move money between accounts, even with your approval.

What worked in the test

Gemini sends Johnson an email summary of each staff meeting, so Duet linked a remark from a meeting to a Slack message. It also worked out who her boss is from a shared document for their weekly one-to-one.

When she planned the next few weeks, Duet noticed that her time off left only a couple of working days before a big draft was due. It proposed a fix, offered to update her notes for the editor and drafted the Slack message. The draft copied her habit of writing in lowercase, and Slack labelled it "sent with Claude".

What did not

Johnson had told an interviewee that her story would likely run on Tuesday. Before 9 a.m. that day Duet warned her that the piece still waited for an edit. It missed that the date was loose and that her editor's local time was 7 a.m.

Duet also kept offering to start her drafts, including a draft of the review of itself. Anthropic told users that Duet may misread a thread or misjudge urgency, and that it is still testing when the agent should message you.

Other companies push the same idea: Instinct, Meta's Muse and OpenAI's Dots, which we compared in our piece on AI agents. Duet is the most focused on work.

If you try Duet, connect the apps where your deadlines live and keep the approval step on for every message it drafts. In Johnson's test it found gaps between inbox and calendar better than it judged how urgent they were.

Anthropic has not said when Duet will leave research preview or reach users outside the US. This text is for information only.