Key Notes
- Claude Duet is a proactive cloud-based assistant that continues tasks after the app closes and uses memory, skills and connected services from an existing Claude account.
- The US research preview is opening to Pro and Max subscribers through a waitlist, with background work counting toward shared subscription limits.
- Anthropic says sensitive actions require approval, while the preview can still misread context, misjudge urgency and miss items between account checks.
Anthropic has introduced Claude Duet, a proactive AI assistant that takes assignments through text or voice, works on its own cloud computer and contacts users when a task is finished or a decision needs their attention. Announced on October 21, the research preview is opening through a waitlist for Claude Pro and Max subscribers in the United States.
Duet carries over memory, skills and connected apps from an existing Claude account, according to its product page. It is available through separate iOS and Android apps and on the web. The pitch is continuity: users can hand over work without keeping a chat window open or rebuilding their account context for each assignment.
A Cloud Computer That Keeps the Assignment Moving
In its announcement, Anthropic says Duet has a virtual machine with its own browser, allowing work to continue after someone closes the app or switches off their device. The assistant decides which connected tools to use and returns completed work for review, rather than requiring the user to supervise every intermediate step.
The company’s examples focus on professionals and small businesses: preparing meeting briefs, building dashboards, drafting invoice reminders and assembling presentations. Duet can also send a morning summary of items that need attention. Existing account context is intended to make those tasks easier to delegate, although the usefulness of any result still depends on the information and services the user has connected.
That changes the role of the conversation. A message can become the start of an ongoing assignment, with the assistant returning later with an output or a question. The practical test is whether that handoff saves more attention than reviewing the work consumes.
Opus 5.5, With Shared Subscription Limits
Anthropic’s help guide confirms that Duet runs on Claude Opus 5.5 and does not currently offer a model selector. It determines how much effort an assignment needs. Access is being granted in waves, so joining the waitlist does not guarantee an immediate invitation.
The research preview is included with Pro and Max at no additional charge, but it does not come with an unlimited separate workload allowance. Duet uses the same subscription limits as the rest of Claude, including session and weekly limits. Background check-ins count too, and a substantial assignment can consume more usage than a short chat.
Team and Enterprise plans are not included in the initial preview. The paid-plan requirement, US restriction and invitation process also make this a limited launch, rather than a new assistant available to every Claude user. For subscribers, the relevant cost is both the plan they already pay for and the capacity background work takes from it.
Proactive Work Still Has Approval Gates
Anthropic says Duet can read, draft and prepare work independently, but must show the proposed action and wait for approval before sending messages, making payments, signing documents or deleting anything. The announcement also describes safeguards intended to keep passwords and sign-in tokens out of the model’s direct view and to reduce leaks of sensitive information.
Those are the company’s stated controls, not a guarantee that every task will be handled correctly. The preview can misread a conversation or judge urgency badly. Users cannot currently watch its cloud browser or take over that browser, which limits how closely they can inspect a difficult interaction as it happens.
Duet also checks connected accounts at intervals instead of monitoring them continuously. That helps contain background usage, but an urgent item can arrive between checks. The promise of an assistant that keeps working therefore needs to be separated from an expectation that it notices every development immediately.
Duet Joins the Personal AI Agent Race
The launch puts Anthropic in the same broad market as OpenAI’s Dots: persistent assistants designed to carry an assignment across sessions and return with results or requests for input. That is a change from evaluating an AI mainly by the quality of a single answer.
OpenClaw has also been pushing ongoing agent workflows, with recent updates aimed at simpler setup and shared sessions. Duet’s approach centers on an existing Claude account and a managed cloud environment. That makes the products relevant comparisons, without establishing that they offer identical capabilities or that one is already more reliable.
For Anthropic, the opportunity is to turn the context people have already built inside Claude into useful work between conversations. The preview will test a more demanding proposition than producing a good draft: choosing when to act, when to stay quiet and when a human decision is necessary.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
