Claude Duet Debuts: Anthropic's Always-On Agent Targets Meta Muse and OpenAI Dots

News Summary
Anthropic announced Claude Duet on October 21, 2026, a research preview of a proactive AI agent that works in the background across a user's email, calendar and connected apps. Coverage describes it as Anthropic's answer to Meta's Muse and OpenAI's Dots, and it is aimed at professionals and small business owners who have more to track than they can handle. Access starts with Claude Pro and Max subscribers in the United States through a waitlist. Reports say it is powered by Claude Opus 5.5.
What Claude Duet Is
Claude Duet is an always-on agent rather than a chatbot that waits for questions. Users can type or record voice messages. A main conversation thread persists, and separate channels for different topics share the same memory. It reuses the memories, skills and connectors from regular Claude and can search past conversations. It runs on its own cloud-based computer, so it keeps working after the app is closed.
Key Features
Each day, Duet reviews email and calendar, flags priorities, starts first steps such as drafting replies, and follows up on quiet threads. This is the "morning brief." For larger jobs it picks the apps it needs and returns editable files such as research briefs, financial models, slide decks or dashboards in a Make tab. A Tasks tab shows ongoing work and what needs attention, and only the user can mark a task finished. It supports plugins, including one for small businesses, and can use its own cloud browser for sites without connectors. Payment integrations with Stripe, Link and Shop Pay let it check unpaid invoices, draft reminders and make purchases with the user's approval.
Safety and Approvals
Anthropic says Duet is designed to ask before it sends a message, makes a payment, signs a document or deletes anything. Reports also mention a dedicated classifier and safety system prompt, and that Duet does not see passwords or logins when given access to accounts.
Availability and Cost
The preview is limited to Pro and Max subscribers in the US, with invitations sent in waves from a waitlist. There is no extra charge, but usage, including background work, counts toward plan limits, which may run out sooner than with chat alone. Team and Enterprise plans are not included yet, though Anthropic plans to add them. Outlets differ on platforms: 9to5Mac reports standalone iOS and Android apps plus the web at claude.ai, while MacRumors describes an iOS app. Outlets also differ on whether the announcement fell on a Tuesday or Wednesday, but all agree on October 21, 2026. No time of day was given in the coverage we found.
Competitive Context
Meta's Muse is a personal agent built on its Muse Spark model family, introduced in April 2026, and OpenAI's Dots arrived a few weeks before Duet. All three aim at everyday follow-up work such as inbox, plans, forms and purchases. Duet's differences are its reuse of existing Claude memory and connectors and its cloud computer for longer tasks.
Why It Matters for Learners
Proactive agents shift AI from answering prompts to monitoring context and acting with permission. Studying how Duet handles approvals, memory and background compute is a useful case study in agent design and human oversight.
Sources and Caveats
This report draws on 9to5Mac, 9to5Google, MacRumors, CNBC and Seeking Alpha coverage. The Verge article could not be retrieved, and CNBC could not be read in full. Details such as the Opus 5.5 model and platform list come from secondary reports and should be checked against Anthropic's announcement.
This article was compiled by the AIBARS editorial team with AI assistance. AI can make mistakes, so please check the original source for anything important. Spotted an error? Email [email protected].