Anthropic's Agent Moves Deeper Into Users' Workdays

Anthropic introduced Claude Duet on Wednesday, Oct. 21. The AI agent can read users’ email, calendars and business software to deliver briefings and flag items that need attention. It moves Anthropic’s chatbot beyond answering questions and into handling parts of users’ working lives. It’s an early test, but the design places a corporate system between people and the tools they use to manage their work.

A Workday Routed Through an Agent

Claude Duet sends users a morning briefing drawn from their email, calendar and business software. Every few hours, it checks those accounts and contacts users if it detects something that may need attention. The agent isn’t limited to responding when someone asks a question; it’s designed to monitor connected work accounts and decide when to send an alert.

The agent connects to more than 6,000 outside apps, including Canva and DocuSign. Those connections sit at the center of the product Anthropic introduced: Claude Duet is meant to work across services people already use, rather than only inside a chatbot conversation. Anthropic is offering an assistant that can take actions during a workday, as well as sum it up.

Anthropic says users will see exactly what the agent plans to do before it sends a message or makes a payment. That preview gives users a chance to review proposed actions. But some actions remain out of reach, even with a user’s approval: Claude Duet can’t trade stocks or move money between accounts.

Purchases, Permissions and Limits

Claude Duet can make purchases through Link, Stripe’s digital wallet, or Shopify’s Shop Pay. Each purchase requires user approval. Anthropic has placed a human approval step between the agent and a completed purchase. That’s a limit on the agent’s autonomy, not a separation from the commercial systems it can access.

Anthropic also introduced a filter designed to stop Claude Duet from sharing credit-card numbers or government ID numbers. The company describes this as a safeguard, while the agent’s broader design still relies on connections to email, calendars and business software. Anthropic says users will see what it plans to do before a message or payment. The source account does not describe a grassroots or community-run alternative to the product.

A Corporate Race for the Assistant Role

Anthropic said Claude Duet competes with OpenAI’s dots, which debuted one month earlier, as well as Meta’s Muse, Google’s Gemini Spark and agents from start-up Instinct. Instinct raised $1 billion in its latest funding round. These competitors and that funding figure place the launch in a race among companies to build agents that handle tasks across users’ digital work lives.

Subscribers to Anthropic’s Pro and Max plans in the United States can join a wait-list to use Claude Duet at no additional cost beyond their subscription fee. That’s the access Anthropic announced: a wait-list for existing paid-plan subscribers, not a product available to everyone. Users now face a new set of proposed actions and approvals, while Anthropic and its competitors build the systems that will carry them out.