Does Claude Duet cost extra?
Per the report there is no separate price. Duet uses your existing Claude plan limits, and its background checks count toward them.
The report does not say how much the background checks actually consume, so whether you need extra allowance is something to find out by watching usage once you have access.
Can I use it outside the US or on a free plan?
Not for now. The report says it is limited to Pro and Max subscribers in the United States, and you must join a waitlist first.
The report gives no timeline for other regions or plans, so go by Anthropic's official announcements.
Could it pay or send mail without my knowing?
Per the report, Duet asks for your approval before sending messages, making payments, signing documents or deleting files.
That is Anthropic's design description, and actual behavior is yours to verify: start with a few connected apps and check that each approval request it raises is what you expect.
How is it different from ordinary scheduled automation?
Scheduled automation does only what you wrote in advance, while Duet's selling point is deciding for itself what is worth flagging. That raises convenience and also makes behavior harder to predict.
So when evaluating it, look at what it over-flags as well as what it misses, since every unnecessary check or alert spends your allowance and attention.
Anthropic launched Claude Duet on October 21, 2026, positioned as a proactive AI Agent that reaches out before you ask. MacRumors reports it is a standalone product in research preview, placed against Meta's Muse and OpenAI's Dots. What follows comes mainly from that report; I did not read the full text of Anthropic's own announcement, so check details against the official description.
Per the report, Duet connects to your apps and flags items needing attention, sends a short morning brief on important items, and checks connected apps every few hours to send alerts when something needs you. It completes tasks in the cloud on a virtual machine with its own browser, without access to your local computer. The scope runs from simple admin tasks to more complex work such as building web apps. It is built on Opus 5.5, and Anthropic says the intended users are professionals and small business owners who have too much to track.
While working independently, Duet asks for your approval before sending messages, making payments, signing documents or deleting files. It draws on the memory, skills and connected apps already in your Claude account, so setup is minimal, and all interaction happens in a single ongoing thread. On privacy, the report says it does not see passwords or logins when given account access, and that a dedicated classifier and safety System Prompt screen each message to guard against leaks of sensitive information. These are Anthropic's claims, and there is no third-party verification of how well the safeguards work.
For now it is waitlist-only and limited to Pro and Max subscribers in the United States, on the web and through a standalone iOS app. The report lists no separate price: Duet uses your existing Claude plan limits, and its background checks count toward them. The report does not say how much the background checks actually consume, which is the biggest unknown, because checking every few hours means allowance is being used even when you are not.
If you have a Pro or Max subscription and join the waitlist, the thing to think through in advance is allowance: with an agent that runs on its own, usage is no longer decided by each question you ask. In your first week with access, check usage once a day, note what the morning brief and the background checks each use, and only then decide which apps to connect. Keep permissions for outbound messages and payments at approval-required rather than loosening them for convenience. Until Anthropic explains background usage, treat it as a subscription feature with an ongoing cost, not a free extra.