Claude Duet explained: a proactive Claude that keeps working on its own VM

DARKMATTER @darkmatrxOriginally posted 2026.10.22

Open original on X

Claude Duet: the first Claude that starts the conversation Every AI assistant so far has worked the same way. You open it, you type, it answers. Close the tab and it stops. Claude Duet flips that. Anthropic opened a research preview on October 21, and the pitch is simple: Claude comes to you with what needs your attention, does the tasks you hand it, and keeps going after you walk away. What it actually does Each morning it sends a short brief built from your connected apps. During the day it checks those accounts every few hours and messages you only when something needs a decision. Everything happens in one ongoing thread. You can ask for something quick or hand it a project that runs for weeks, by text or voice. The work runs on Duet's own virtual machine with its own browser. Your phone and laptop don't need to stay on. You get a notification when a task is done or needs review. Anthropic lists the kinds of tasks it's built for: scheduling meetings, building dashboards, financial and sales forecasts, prototypes and web apps. What it connects to Duet carries over your memory, skills and connected apps from your Claude account. It can use over 6,000 connectors and plugins from the Claude Marketplace, including Gmail, Slack, QuickBooks, Stripe, Gusto, LinkedIn, DocuSign, HubSpot, Canva and Shopify. It can also buy things through Stripe's Link with a virtual card, or check out with Shop Pay. Where it stops This is the part that matters most for an agent with access to your inbox and payments. Duet reads, drafts and prepares on its own. Before it sends a message, makes a payment, signs a document or deletes anything, it shows you the action and waits for approval. Some things are off-limits even with approval: trading securities and moving money between accounts. It never sees your passwords or sign-in tokens directly. A separate classifier blocks sharing card numbers and government IDs. Health and personal finance details are shared only when you explicitly tell it to. What Anthropic admits it can't do yet It can misread a thread or misjudge what's urgent. Personal errands like booking tickets are weaker than work tasks. You can't watch its browser yet. And background checks count against your usage limits, so a busy agent burns through them faster. Who can get it Waitlist only, for US Pro and Max subscribers. It's included in those plans with the same limits. It runs on Claude Opus 5.5 and lives in a separate app on iOS, Android and the web, apart from the main Claude app for now. Why this one is worth watching Most "agent" launches are a chat window with extra steps. Duet is a different shape: a background worker with a fixed approval line. Whether people trust it with their inbox will decide if this becomes the default way to use Claude, or stays a power-user app.

This post quotes another post: x.com/claudeai/status/2112984664256241837