INDEPENDENT JOURNAL

A clearer view of what matters

An independent news site covering AI, technology, economy, society, science, health, sports, culture and world affairs.

AI · 2026-10-24

Anthropic unveils “Claude Duet,” an AI agent that keeps working even when you step away

Anthropic has unveiled “Claude Duet” as a research preview: an AI agent that reviews connected apps such as email and calendars, organizes tasks, and keeps working while the user is away. Sensitive actions like sending messages or making payments require user approval.

Anthropic unveils “Claude Duet,” an AI agent that keeps working even when you step away
AI-generated editorial illustration
AI is moving beyond simply answering questions and is now starting to keep track of tasks and carry them out on the user’s behalf. Anthropic’s newly unveiled “Claude Duet” is a personal AI agent that checks apps such as email and calendars, organizes important work, and continues operating on a separate virtual computer even when the user steps away.

According to Newsway, Anthropic released the research preview of Claude Duet on October 22, local time. The system is built on Claude Opus 5.5 and is designed to help professionals and small business owners handle multiple tasks. For now, users in the U.S. on the Pro and Max plans can apply through a waitlist, and it is available on iPhone, Android, and the web.

In the morning, Claude Duet reviews connected apps and provides a summary report of important information. Even during the workday, it checks email and schedules every few hours. If something requires the user’s attention, it sends an alert, and it also supports everyday tasks such as rescheduling meetings or drafting email replies. It has also been introduced as capable of taking on more complex work, including building dashboards, analyzing financial and sales forecasts, and developing web applications.

One of its standout features is that it can choose the apps and tools it needs on its own and keep working even after the user closes the app or leaves their desk. When a task is finished or further judgment is needed, it sends a notification and hands over the results for review. It also uses memory stored in the Claude account, technical settings, and connected apps. Anthropic cited examples such as checking messages in Gmail and Slack, reviewing unpaid balances in Stripe and drafting reminder emails, writing job postings in LinkedIn, and managing Shopify stores. The company says the Claude marketplace offers more than 6,000 connected features and plugins.

As the scope of delegated work expands, safety measures become even more important. Claude Duet is designed to request user approval before carrying out sensitive actions such as sending messages, making payments, signing documents, or deleting files. It does not allow securities trading or transfers between accounts, regardless of approval, and it also blocks requests to share sensitive information externally, including credit card numbers and government-issued ID numbers.

That said, because it is still at the research preview stage, there are limitations. Anthropic says the system may misread conversational context or fail to judge how urgent an issue is, and it could miss emergencies that arise between scheduled check-ins. For now, then, it may be more realistic to start with tasks whose results can be reviewed rather than handing over everything at once.

The release of Claude Duet shows that AI agents are evolving beyond simple question-and-answer tools and moving toward a role in ongoing workflow. At the same time, because they connect to email, calendars, and payment services, approval steps that preserve user control and ways to handle errors are likely to become increasingly important. This summary is based on a Newsway report.

Related search terms

#Claude Duet#Anthropic Claude Duet#AI agent research preview#email and calendar AI assistant#workflow automation AI#Claude Opus 5.5#AI assistant with user approval
← Back to articles