Anthropic has its own answer to the AI agent wave, and it arrived Wednesday. The company opened a research preview of Claude Duet, a standalone app that turns Claude into a proactive agent aimed at professionals and small-business owners. “It comes to you with what needs your attention, completes tasks you ask of it, keeps working even when you step away, and checks with you before it acts on your behalf,” Anthropic said of the new app, as reported by Engadget. Access runs through a waitlist, starting with Pro and Max subscribers in the US.
Here’s the flip. Instead of waiting for a prompt, Duet starts each day by sending a short briefing of the most important items across your connected apps, then checks those accounts every few hours and messages you only when something actually needs you. The agent runs on its own virtual machine with its own browser, so your laptop or phone doesn’t have to stay on while it works. And that background activity isn’t free: Anthropic says the checks count toward the same usage limits as the rest of your Claude plan, because nonstop monitoring would burn through a plan’s limits far faster.
Duet is built around Claude Opus 5.5, one of Anthropic’s latest models. It carries over the memory, skills, and connected apps from your Claude account, and you talk to it by typing or voice in one ongoing thread. When it finishes a task — or needs a decision — it sends a notification.
What Claude Duet can actually do
The connector list is the part small-business owners should read first. Anthropic says Duet can use more than 6,000 connectors and plugins in the Claude Marketplace, and its own examples are telling: checking Intuit QuickBooks or Stripe for unpaid invoices and drafting reminders, preparing and running payroll in Gusto, drafting follow-up emails from HubSpot contact and deal history, sending agreements out for signature through DocuSign, and even setting up a Shopify store.
Tasks scale up from the mundane to the heavy. Scheduling meetings is on the simple end; building dashboards across dozens of tools, creating financial or sales forecasts, prototyping, and building web apps sit on the complex end. Logos on the Duet product page include Stripe, 1Password, Shopify, Gmail, Google Calendar, Slack, Intuit QuickBooks, DocuSign, Google Drive, HubSpot, Gusto, Microsoft 365, Notion, Canva, and Ramp — a roster that reads like the back office of a typical small company.
Money moves through familiar rails. Purchases made by the agent go through Stripe’s Link wallet with a virtual card, or Shopify’s Shop Pay, according to coverage of the launch. Anthropic adds that Duet never directly sees passwords or sign-in tokens, and that approvals, logins, and safety checks run inside the user’s own cloud computer rather than on Anthropic’s servers.
The approval rule is the whole point
Anthropic’s biggest design bet is restraint. Before Duet sends a message, makes a payment, signs a document, or deletes anything, it shows you exactly what it’s about to do and waits for your approval. Some actions are off-limits even with approval — trading securities and moving money between accounts can’t be authorized at all. A dedicated classifier is designed to reject requests to share credit card numbers and government IDs, and other sensitive information like health or personal finance details goes out only when you’ve explicitly told Duet to send it.
That posture stands out in a month where the competition has leaned the other way. OpenAI launched its always-on “dots” agents last month to pursue user goals across apps like Slack and Microsoft Teams, and Meta’s Muse AI agent drew millions of downloads and set off a wave of investor interest in autonomous AI services, Reuters reported. Anthropic is deliberately positioning Duet as the agent that asks first.
It also sets the bar for trust in plain terms. Sensitive operations aren’t just confirmed — Anthropic says “approvals, logins, and safety checks run in your own cloud computer, so sensitive details stay private.” For a tool that can see invoices, payroll, and deal pipelines, that architecture choice matters as much as the feature list.
Why this matters
The honest catch is usage. A proactive agent that watches your apps all day creates an awkward math problem: every background check draws from the same Pro or Max cap as your chat window. Anthropic flags this openly, noting that Duet checks accounts every few hours precisely because nonstop checking would chew through limits, and that larger tasks like building a financial model or a deck cost more model turns on top of that. The busiest users, the exact people Duet is built for, are the ones most likely to feel the cap.
Then there’s accuracy. Anthropic says Duet “may misread a thread or miscalibrate what’s urgent,” and that it can miss something that arrives between its periodic check-ins. Reuters added that the agent remains less effective at personal tasks like booking tickets — work, not play, is the home turf here. A research preview label earns the company some slack, but an agent that acts on your behalf lives or dies on judgment calls, and judgment calls are where language models still stumble.
Still, the launch marks a clear shift in how Anthropic sells Claude. The company has spent months releasing agent-shaped products, and Duet bundles them into something a non-technical business owner can download and use. Morning briefing, background worker, approval-gated actions: it’s the agent model with training wheels firmly attached. Whether users embrace an assistant that taps them on the shoulder — or find a second feed of notifications exhausting — is the experiment now running. The waitlist is open, the app is free for US Pro and Max subscribers during the preview, and Anthropic says it will email people as their access is ready.
Sources: Reuters, Engadget, Unite.AI, WebProNews.
FAQ
What is Claude Duet? Claude Duet is Anthropic’s new standalone AI agent app, opened in a research preview on October 21, 2026. Built on Claude Opus 5.5, it proactively surfaces what needs your attention, completes tasks you assign, and runs on iOS, Android, and claude.ai.
Who can try Claude Duet? Access is by waitlist, starting with Pro and Max subscribers in the US. During the research preview it is included in Pro and Max plans and uses the same usage limits as regular Claude.
Is it safe to let Claude Duet act on my accounts? Duet shows exactly what it is about to do before sending a message, making a payment, signing a document, or deleting anything. Trading securities and moving money between accounts are off-limits even with approval, and a classifier blocks sharing of credit card numbers and government IDs.
How is it different from OpenAI’s dots or Meta’s Muse? The pitch is approval-first design rather than autonomy-first. Duet sends a morning briefing from your connected apps, works on its own virtual machine in the background, and checks in only when something needs your attention.