Anthropic is testing a persistent AI agent that can carry out tasks in a cloud-based virtual machine, with user approval required for sensitive actions.
Anthropic has introduced Claude Duet, a research preview of an AI agent designed to keep working in the background and act on behalf of users. The service is being rolled out first to US subscribers on the Pro and Max plans, with access issued gradually from a waiting list. Anthropic has not said when it will become available in Japan.
Unlike the standard Claude app, Claude Duet is offered as a separate product, while still carrying over the user’s saved memory, skills and connections to external services. The company says it can handle tasks such as preparing morning briefings, drafting documents and assembling dashboards, then alert the user when it has finished or needs confirmation. Because it runs inside a cloud-based virtual machine with its own browser, it can continue operating even if the user’s phone or computer is switched off. Anthropic says actions including sending messages, making payments, signing documents and deleting items will require user approval, while some higher-risk activity, such as securities trading or transfers between accounts, is blocked entirely.
The launch reflects a wider shift in the AI sector away from chat interfaces and towards persistent agents that can execute longer workflows with limited supervision. In recent months, other major developers have announced similar tools, including Grok Bot from SpaceXSI, Muse from Meta and dots from OpenAI, underscoring how quickly the market is moving towards cloud-hosted assistants that can search, decide and report back without constant user intervention.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
