Anthropic Launches Claude Duet AI Agent Test

Anthropic introduced Claude Duet on Wednesday, Oct. 21, releasing an AI agent in an early test phase that works as a personal digital assistant. The company says it’s built for work, not personal errands. The launch marks a clear move for Anthropic’s Claude chatbot beyond answering questions and toward handling parts of people’s working lives.

Claude Duet doesn’t wait to be asked. It sends users a morning briefing based on their email, calendar and business software. Then it checks those accounts every few hours, reaching out when it spots something that may need attention. For workers juggling inboxes and meeting schedules, that constant monitoring could save hours. It could also mean a machine reads a great deal of private correspondence on an ongoing basis.

The Competitive Field

Anthropic said Claude Duet competes with OpenAI’s dots, which debuted the previous month. It also goes up against Meta’s Muse, Google’s Gemini Spark, and agents from the start-up Instinct, which raised $1 billion in its latest funding round. The crowded field points to where the industry thinks the money is going: away from chat windows and toward software that acts on a user’s behalf.

The agent connects to more than 6,000 outside apps, including Canva and DocuSign. That reach makes it useful, but it’s also a concern. Every connection opens another door, and the people behind those doors aren’t always the ones who’ll pay the price if something goes wrong.

Guardrails and Purchases

Claude Duet can make purchases through Link, Stripe’s digital wallet, or Shopify’s Shop Pay. Users must approve each one. Anthropic said the agent will show users exactly what it plans to do before it sends a message or makes a payment. That’s a meaningful safeguard. It’s the kind of human checkpoint that matters when software starts spending real money.

Some actions remain off the table. The system blocks stock trading and transfers between accounts, even when a user approves them. Anthropic also added a filter to stop the agent from sharing credit-card numbers or government ID numbers. Those limits reflect a recognition that an assistant with access to financial and identity data needs hard edges, not just good intentions. Whether the filters hold up in real-world use is something the early test phase will have to answer.

Access and Cost

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. The setup gives access first to people already paying for premium service. Anyone without those plans, or outside the United States, has no route to the tool yet. As these agents mature, who gets the productivity gains—and who gets left out—will matter as much as the technology itself.

Why This Matters: AI agents that read private email, calendars and financial accounts move a great deal of personal information into corporate hands. Anthropic says Claude Duet requires user approval for purchases and blocks certain high-risk actions; those protections deserve close scrutiny. But the product depends on continuous access to sensitive data, while the companies building these tools still set most of the rules governing that access. Workers who rely on such systems for their jobs have little say in how companies use their information. As Claude Duet, dots, Muse, Gemini Spark and Instinct’s agents spread across the market, the public will need clear standards for data handling, consent and accountability. Without them, a digital assistant’s convenience could quietly become a question of who controls the most intimate details of ordinary working life.