Skip to main content
Live
Main content

Anthropic launches Claude Duet, a proactive Opus 5.5 agent for Pro and Max users

The research-preview agent runs on its own cloud computer, sends a morning brief, and asks before it sends, pays, signs or deletes anything.

Jaeden Schafer
Editor in Chief · · 5 min read
Anthropic logo

Anthropic launched Claude Duet on Wednesday, a research-preview personal agent that runs on Claude Opus 5.5, works from its own cloud computer, and messages users when something needs their decision. Access is waitlisted for US subscribers on the Pro and Max plans, with invitations going out in waves.

Claude Duet ships as a standalone app for iOS and Android and runs in the browser at claude.ai for invited users. Anthropic says it is not yet available on Team or Enterprise plans but plans to bring it there later. During the preview it carries no extra charge. Its usage, including background work, counts against a subscriber's existing plan limits, and Anthropic warns users may hit those limits sooner than with chat alone. Newly admitted users get one full reset of their five-hour and weekly limits.

The product is built around a daily routine. Each morning Duet reviews a user's email and calendar, highlights priorities, and can start first steps such as drafting replies. Through the day it posts progress updates and follows up on stalled threads, then stays quiet until something needs attention.

Key facts

  • 01Anthropic opened a waitlisted research preview of Claude Duet for US Pro and Max subscribers on October 21, 2026.
  • 02Duet runs on Claude Opus 5.5 on its own cloud computer, so tasks keep running after the app is closed.
  • 03It asks before sending messages, making payments, signing documents or deleting anything.
  • 04There is no extra charge during the preview; usage, including background checks, counts against existing plan limits.
  • 05Team and Enterprise plans are not included yet, and Microsoft 365 is not yet available as a connector.

Larger jobs run in the background. Anthropic's examples include setting up meetings, building dashboards, financial models, decks and briefs, finding unpaid invoices and drafting payment reminders, preparing payroll for review, and getting contracts ready to send. Because the agent runs on its own virtual machine, tasks continue after the app is closed. Finished deliverables land in a Make tab as editable files, and a Tasks tab tracks ongoing work. Only the user can mark a task complete.

Duet carries over memory, skills and connectors from the regular Claude app and can search past conversations. Anthropic's product page says it can reach more than 6,000 connectors and plugins, naming Gmail, Google Calendar, Slack, Shopify, Stripe, QuickBooks, HubSpot and Notion among them. For sites without a connector it uses its own cloud browser, and it supports plugins such as a small-business bundle. Its answers name the email, file or spreadsheet they came from.

Approval gates sit at the center of the design. Duet asks before sending messages, making payments, signing documents or deleting anything. If a user can't approve right away, it finishes the rest of the task and brings the request back later. Purchases through Link and Shop Pay use a one-time card that works only for that transaction, and Anthropic says Duet will not trade securities, move money between accounts, or access someone else's account even with approval.

On security, Anthropic says the agent treats content from emails and web pages as information rather than instructions, to resist prompt injection, while noting that no defense is perfect. Passwords go through a separate form that sends them straight to the browser, keeping them out of the chat, and a new classifier and system prompt are meant to prevent leaks of sensitive information. Anthropic said it would publish more detail on its safety protections in the coming days.

The preview has stated limits. Users can't choose the model, since Duet runs on Opus 5.5 and sets its own effort level. It can't make phone calls or send texts, users can't watch or take over its browser, and Anthropic's support page says Microsoft 365 isn't available yet while Slack works with limits. Duet checks connected accounts at intervals of a few hours rather than continuously, which Anthropic says keeps plan limits from draining faster, so urgent items can arrive between check-ins.

Related · from this week
Anthropic relaunches Claude Code Projects to orchestrate multiple agents in the cloud
Jaeden Schafer · 4 min read →

Anthropic's FAQ also cautions that Duet may misread a thread or miscalibrate what's urgent, and the company says it is still testing the agent's judgment about when to message users. Users will be able to set how often Duet reaches out and which topics it covers.

Claude Duet arrives in a category where Meta Muse and OpenAI's Dots are already available as personal agents. OpenAI currently limits Dots to its Pro tier, with plans starting at $100 a month, while Anthropic says $20-a-month Claude Pro subscribers will get Duet when it reaches general availability.

ShareXLinkedInEmail
AI Box

Every AI model. One chat.

The latest models from ChatGPT, Claude, Gemini, Sora, ElevenLabs — 80+ models in a single chat. Compare answers side by side. Pick the best one every time.

  • ChatGPT, Claude, Gemini, Grok, DeepSeek — in one chat
  • Generate images & video with Sora, Veo, Ideogram
  • Compare any two models side by side
  • From $8.99/mo · 80+ models, all included
Try AI Boxaibox.ai
Trusted by 3,000+ teams
Got a tip?

Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.

Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.

AI Box Daily briefingFree · Daily · No fluff

Stay ahead of everyone in AI.

The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.

Loved by 10,000+ AI professionals
Free forever. Unsubscribe with one click.

The briefing read inside teams at

Keep reading

More from Tools

Anthropic logo
Tools

Anthropic relaunches Claude Code Projects to orchestrate multiple agents in the cloud

The revamped Projects feature runs parallel Claude Code threads on branched repos, with a coordinator resolving conflicts as pull requests.

Jaeden Schafer4 min read
Anthropic logo
Tools

Anthropic flips Claude Code into auto mode by default

Starting August 14, Pro, Max, and Team accounts get an agent that stops asking permission at every step.

Jaeden Schafer4 min read
Hark previews Handoff, a browser-use agent for real-world tasks
Tools

Hark previews Handoff, a browser-use agent for real-world tasks

The $700M-funded startup claims its post-trained model beats GPT 5.5 and Opus 4.8 on cost and speed for browser automation.

Jaeden Schafer4 min read