Compiled from reports by MacRumors and Engadget. See the original articles linked below.
Anthropic has introduced Claude Duet, an AI agent that is meant to start the conversation itself instead of waiting for a prompt. MacRumors and Engadget, both writing on October 21, 2026, said the product is a research preview and positioned against Meta's Muse and OpenAI's Dots.
What the agent does
MacRumors reported that Duet links to a person's apps, points out items that need attention, and can finish tasks in the cloud using its own virtual machine and browser. Because of that, it does not need access to a local computer. According to MacRumors, Anthropic built the product for professionals and small business owners who have many things to keep track of.
Engadget wrote that Duet produces a short daily briefing of action items drawn from connected apps and otherwise messages users only when something calls for it. The outlet listed scheduling meetings, building dashboards, forecasting and prototyping among the tasks it handles. MacRumors added that the agent sends a brief each morning and checks the connected apps every few hours.
Both outlets said the agent runs on Claude Opus 5.5. MacRumors said it draws on the memory, skills and connected apps from a user's existing Claude account, with all interaction held in a single continuing thread. Engadget said Duet runs on its own virtual machine, so no Claude session has to be open.
Safeguards
MacRumors reported that Duet asks for approval before sending messages, making payments, signing documents or deleting files. It also said Anthropic states that the agent never sees passwords or logins, and that a dedicated classifier and system prompt are intended to prevent sensitive information from leaking.
Engadget gave a similar account of the safeguards. It said Duet does not directly see passwords or sign-in tokens, and that health or financial information is shared only when the user explicitly asks. The outlet added that Anthropic has promised to publish more safety details.
Access and limits
MacRumors described the preview as a waitlist open to Pro and Max subscribers in the United States, available on the web and in an iOS app. Engadget described the waitlist as limited to US Claude subscribers and said Anthropic is releasing standalone Android and iOS apps. The two outlets therefore differ slightly on the platforms mentioned.
On usage, MacRumors said the background checks count against the limits of a subscriber's existing Claude plan. Engadget said Duet shares usage with other Claude tools, that checking accounts only every few hours is meant to conserve those limits, and that users will be able to choose how often Duet contacts them and on which topics.
Engadget also reported that, according to Anthropic, the $20 per month Pro plan will include access once the product reaches general availability. The outlet noted that OpenAI's Dots are currently restricted to that company's Pro tier, which it said starts at $100 per month. The sources reviewed did not give a date for general availability.
Engadget closed its piece by observing that the usefulness of Duet will depend on how much users trust Anthropic with their digital lives.
Sources
- Anthropic Launches Claude Duet, a Proactive AI Agent (MacRumors, 2026-10-21)
- Claude Duet is Anthropic's answer to Meta Muse and OpenAI Dots (Engadget, 2026-10-21)
This article was compiled with AI assistance from the reports listed above and reviewed by GMA before publishing.