Claude Duet works on its own cloud computer, but joining the waitlist takes a form
You message it a task, it works on a cloud computer with your connected apps, sends a morning brief, and asks before it sends messages or pays. To get in, you type your details into a waitlist form.
Claude Duet is a research preview for US Pro and Max subscribers, powered by Opus 5.5. It runs larger tasks, such as scheduling meetings or building dashboards, on its own cloud computer, and reports back when done.
The first complaint in the thread: a lab with frontier coding agents could let you ask Claude for a spot instead of retyping your details.
What Duet does when you're not looking
Claude Duet is a proactive agent. You send it tasks by message; it works on its own cloud computer using the apps you connect, from a pool of more than 6,000 connectors and plugins. It sends a morning brief of what needs attention, drafts next steps and checks in when a task finishes or needs a decision. Your Claude memory and skills carry over, and it asks for approval before sensitive actions such as sending messages or paying. You can watch what it is doing and steer it.
It is a research preview for Pro and Max subscribers in the US only, on standalone iOS and Android apps or at claude.ai. Team and Enterprise plans are not included, and background work counts against plan usage limits. The page gives no launch date or waitlist size.
A waitlist Claude cannot sign you up for
"Signing up for this waitlist should be something you can request from your Claude chat session." — a commenter
The thread's real benchmark was a rival. A user of ChatGPT Dot, OpenAI's equivalent, reported:
"In the past month since Dot came out, I've literally bought a car with almost the entire process handled by Dot"
That is the yardstick Duet will be held to, and Anthropic's page has no end-to-end errand like it: the dashboard figures on the page, such as $528K in weekly revenue, are labeled as illustrative examples.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
PHP patched two unauthenticated RCE bugs after the CVE IDs leaked two days early
A request that hits memory_limit at the wrong moment can corrupt memory. A fatal error can leave a use-after-free for the next request on the same worker. Both lead to remote code execution.