What happened
On 21 October 2026 Anthropic introduced Claude Duet, which it describes as a proactive agent: instead of waiting for a prompt, it looks at your connected apps, prepares morning briefs, works on tasks while you are away and asks for approval when needed. It is a research preview offered through a waitlist, starting with Pro and Max subscribers in the US. Anthropic makes Claude, the AI tool that helped draft this article, so please read this as a company describing its own product.
Key details
All details below are from Anthropic’s announcement and product page, checked on 22 October 2026. Press coverage dates the launch the same way.
- Access: a waitlist, starting with US Pro and Max subscribers. Anthropic says it will email people when Claude Duet is ready for them. The FAQ says Team and Enterprise plans are “not included yet”.
- Price and limits: no separate fee. Claude Duet is included in Pro and Max and “uses the same usage limits as the rest of Claude”. Anthropic says its background checks count toward those limits, and that larger jobs such as building a financial model or a deck need more model turns and background work.
- Model and where it runs: it is powered by Claude Opus 5.5, in a standalone iOS and Android app and on claude.ai.
- How it works: it checks connected accounts “every few hours”. Anthropic says it carries over Claude memory, skills and apps, and that it can use “over 6,000 connectors and plugins in the Claude Marketplace”. That number is a Marketplace total, not a count of integrations built for Duet specifically, and we could not verify it independently.
- Approvals: before it sends a message, makes a payment, signs a document or deletes anything, Claude Duet shows what it is about to do and waits for approval. Anthropic lists a few actions as off-limits even with approval, such as trading securities or moving money between accounts.
- Sensitive data: Anthropic says a dedicated classifier and safety prompt are designed to refuse requests to share the strictest tier of information, including credit card numbers and government ID. It says the agent never directly sees passwords or sign-in tokens, and that red teams and an external firm tested it and it is covered by Anthropic’s public bug bounty.
- Payments: Stripe Link virtual cards and Shopify Shop Pay are named. Sign-in through saved 1Password logins is described as coming soon.
- Limits Anthropic states: it cannot yet show you its browser; personal errands such as booking tickets are “not quite as strong”; it may misread a thread, miscalibrate what is urgent, or miss something that arrives between check-ins.
- Still to come: Anthropic says it will share more about how Claude Duet is designed for safety “in the coming days”.
Press coverage places it alongside similar proactive assistants from OpenAI and Meta; those comparisons are the outlets’ framing, not Anthropic’s claim.
Why it matters
Claude Duet moves Anthropic’s consumer products from answering requests to monitoring accounts and acting on them, which raises the practical stakes of the permission and approval design. The approval step before messages, payments, signatures and deletions is the main safeguard Anthropic describes, and its effectiveness will depend on how people use it in practice; no independent evaluation has been published yet.
For subscribers, the immediate facts are narrower: it is US-only, waitlisted, and draws on the same usage limits as chat, so frequent background checks could use up allowance faster. Anthropic’s promised safety write-up is the next thing to read.
Disclosure: Claude, made by Anthropic, was one of the AI tools used to research and draft this article, and this article is about Anthropic. Figures and claims are Anthropic’s own and have not been independently verified.
Sources
- Claude Duet (announcement) Primary source , Anthropic
- Claude Duet (product page and FAQ) Primary source , Anthropic
- Anthropic takes on ChatGPT Dots and Meta Muse with Claude Duet 9to5Mac
- Claude Duet app is Anthropic's bet on agents TechCrunch