Claude Duet: Anthropic Launches a Proactive AI Agent Preview

Claude Duet, a new personal AI agent from Anthropic, is now rolling out by waitlist to U.S. subscribers on the company’s Pro and Max plans. Announced Oct. 21, 2026, the research preview is built to work on its own: it sends a morning briefing, watches connected apps like Gmail, Slack and QuickBooks, and handles tasks in the background, then asks for approval before it sends, pays, signs or deletes anything.

The launch puts the San Francisco AI company in direct competition with Meta’s Muse and OpenAI’s Dots agents, CNBC reported, in what has become the industry’s newest race: software that does work for you instead of waiting for a prompt.

What Claude Duet Does

Anthropic describes Claude Duet as a “proactive agent” for 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,” the company wrote in its launch post.

Key features, according to Anthropic’s announcement and help center:

• Daily brief: each morning it scans email and calendar, picks the most important items and can start drafting replies for review.
• Background work: it runs on a cloud virtual machine with its own browser, so tasks continue after you close the app or turn off your phone.
• Bigger jobs: it can build dashboards, financial or sales forecasts, slide decks, research briefs and prototype web apps, with results saved in a “Make” tab.
• Connectors: it can tap more than 6,000 connectors and plugins in the Claude Marketplace, including Stripe, Gusto, HubSpot, Docusign, Canva and Shopify.
• Purchases: with approval, it can buy items through Stripe’s Link wallet or Shopify’s Shop Pay using a one-time virtual card.

It runs on Anthropic’s Claude Opus 5.5 model, and users cannot choose a different one. It is available as a standalone iOS and Android app and on claude.ai, separate from the main Claude app.

Safety Limits and Approvals

Anthropic built hard limits into the preview. Trading securities, moving money between accounts and accessing someone else’s account are blocked even if the user approves. It does not make phone calls or send text messages. Anthropic says the agent never directly sees passwords or sign-in tokens, and that a new classifier blocks requests to share the most sensitive information, such as credit card numbers and government ID numbers. Health and personal finance details go out only when the user explicitly says so.

The company also acknowledged the risk of prompt injection, in which instructions hidden in an email or web page try to steer the agent. Anthropic’s help page says Claude Duet is built to treat what it reads as information rather than commands, “but no defense is perfect,” which is one reason consequential actions need approval. The company said red teams and an outside firm penetration-tested the product, and it is covered by Anthropic’s public bug bounty.

Anthropic is candid that the preview can misfire. It “might misread a thread or miscalibrate what’s urgent,” the company said, and personal errands like booking tickets are not “quite as strong” as work tasks. It checks connected accounts every few hours rather than continuously, so urgent items can arrive between checks.

Pricing, Usage Limits and Who Can Join

During the preview, Claude Duet costs nothing extra for Pro and Max subscribers. It draws on the same usage limits as regular Claude chats, and its background checks and morning briefings count against those limits. To offset that, Anthropic gives new Duet users one full reset of their five-hour and weekly limits, usable until Nov. 30, 2026. Team and Enterprise plans are not included yet, and Microsoft 365 is not yet supported.

Invitations go out in waves. Access is tied to the email address used on the waitlist, which must match the user’s Claude account.

The Race for Personal AI Agents

Claude Duet arrives as rivals report early traction. Meta’s Muse launched last month and rose to the top of Apple’s App Store, CNBC reported, and Meta CEO Mark Zuckerberg told analysts on Oct. 21 that more than 10 million people in the U.S. and Canada have set up Muse agents. OpenAI introduced Dots, its “always-on agents,” at its developer conference. Muse has already run into friction with retailers: Amazon blocked Meta’s Muse shopping agent in September.

For Anthropic, the product lands during a closely watched stretch. An S-1 draft reported this month showed $16.3 billion in revenue and an $80.7 billion loss, and the company recently committed 5% of its shares to a public benefit fund. Small businesses in Greater Boston that already run on QuickBooks, Gusto or Shopify are the kind of users Anthropic says it built Duet for.

Sources

• Anthropic: Introducing Claude Duet
• Claude Help Center: Get started with Claude Duet
• CNBC: Anthropic announces preview of Claude Duet, a personal AI agent to take on Muse, Dots