Anthropic introduced Claude Duet on Wednesday, Oct. 21: a proactive AI agent that checks your connected apps, sends a short briefing every morning and finishes tasks in the cloud. It is a research preview, open by waitlist to Claude Pro and Max subscribers in the US.
What is Claude Duet?
Regular Claude waits for you to open a chat. Duet comes to you. MacRumors describes it as an agent that reaches out before you ask: it connects to your apps, notices what needs attention and suggests what to do next. It is Anthropic’s answer to Meta’s Muse and OpenAI’s Dots, which follow the same always-on idea.
Every morning, Duet sends a brief summary of the most important items across your connected services. After that, it only pings you when something needs action. You hand it tasks by text or voice, and everything lives in one continuous conversation instead of a new chat for each request.
It is built on Opus 5.5 and reuses the memory, skills and connected apps from your Claude account. According to The Decoder, the target users are professionals and small business owners who have a lot to keep track of.
How does it work when your laptop is closed?
Duet runs on a cloud virtual machine with its own browser, so your laptop or phone does not need to stay on. The work ranges from routine admin, like sorting email and to-dos, to bigger jobs such as building a web app.
It plugs into services through more than 6,000 connectors and plugins from the Claude Marketplace, including Gmail, Slack, Stripe, Docusign and Shopify, per The Decoder. For purchases it uses Stripe Link with a virtual card, so your real card number is never shared.
To protect your plan’s usage limits, it checks connected accounts every few hours instead of constantly. That matters: a message that arrives between two checks can sit unnoticed until the next round.
Can it spend money or send emails on its own?
Not without your approval. Duet reads, drafts and prepares actions by itself, but before it sends messages, makes payments, signs documents or deletes anything, it shows you the plan and waits for a yes. Securities trading and transfers between accounts are blocked even with approval.
Anthropic says the agent never directly sees passwords or login tokens. The company also cites defenses against prompt injection (hidden text in an email or web page that tries to give the agent orders), red-team testing and testing by an outside firm. A dedicated classifier watches outgoing messages to stop sensitive data from leaking. Requests to share card numbers or ID documents should be refused, and health or personal finance details only go through if you ask explicitly.
What are the limits of the first version?
The early version can misread a conversation, misjudge how urgent something is and miss messages that land between checks. It was built mainly for work, so it does worse on personal chores like booking tickets.
It also shares usage limits with other Claude products, and background checks count against them. If you already hit your plan’s cap, expect Duet to eat into it.
Who can use it and what does it cost?
There is no separate price. Duet is tied to Pro and Max, and access is by waitlist for US subscribers. Team and Enterprise plans are not included for now. MacRumors says you can reach it on the web or through a standalone iPhone app called Duet, separate from the Claude app. How the two will merge is still open.
Is it worth waiting for?
If your day is a pile of email, payments and documents spread across apps, a ready-made morning briefing beats opening five apps. The catch is trust. An agent acting for you has to be right most of the time and ask permission whenever a mistake would be expensive, and Duet’s required approvals for sensitive actions point that way.
Keep expectations grounded, too. As we explain in our piece on what a language model is, these systems still make things up and misread context, which is why human review stays central. The false police tip involving Claude is a reminder that safety remains the sensitive topic. With Meta and OpenAI already fielding proactive agents, this looks like the next big fight among AI companies. The real test for Duet will be simple: does the morning briefing save you time, or does it just add one more notification?
