Anthropic has announced Claude Duet, a proactive AI agent entering a waitlist-based research preview for eligible Pro and Max subscribers in the United States. Subscribers can use the service on the web and through a dedicated iOS app to monitor connected applications and surface work needing attention.
Built on Anthropic’s Opus 5.5 model, Duet can deliver a daily briefing and run additional checks as the day progresses. It maintains account-linked memory, skills and connected apps in one continuing conversation, as first reported by Engadget. Anthropic aims the tool at professionals and small-business owners managing a steady stream of ongoing tasks.
Duet runs cloud-based tasks in a virtual browser
The agent operates through a virtual machine that includes its own web browser, allowing it to carry out cloud-based work without direct access to a user’s local computer. Anthropic says the system can handle administrative work and larger projects, including creating web applications.
Duet asks for approval before it sends messages, makes payments, signs documents or deletes files. Anthropic also says it does not receive users’ passwords or login credentials. A classifier and safety system prompt aim to detect attempts to expose sensitive information.
Existing plan limits will govern background work
The launch puts Anthropic alongside Meta Muse and OpenAI Dots in the race to make AI systems take initiative between user prompts. That matters for subscribers whose work spans email, documents and web services: background checks may save time, while they also draw from the usage limits on an existing Claude plan.
Anthropic has not said when it will open the waitlist to every eligible US subscriber, whether it will keep Duet as a separate product after the research preview, or whether it will introduce pricing or usage allowances beyond current Claude plan limits.
This article was produced with AI assistance from multi-source reporting and is published under our editorial standards.
