Anthropic launches Claude Duet AI agent for small businesses

Anthropic announced the launch of Claude Duet, an AI‑driven agent aimed at professionals and small‑business owners. The service is being released as a research preview and is available as a standalone app for iOS and Android as well as through Claude on the web. Access requires a Pro or Max subscription in the United States and enrollment on a waitlist.

What Claude Duet Offers

Claude Duet builds on Anthropic’s Claude Opus 5.5 model and is positioned as a productivity‑focused assistant. Each morning the agent delivers a concise briefing that highlights the most important items across a user’s connected applications. Throughout the day it monitors those apps and only notifies the user when something requires attention. Users can assign a wide range of tasks, from simple scheduling to more involved work such as building dashboards, generating financial or sales forecasts, prototyping, and even creating web applications. The agent accesses a virtual machine that includes a browser, allowing it to retrieve information and complete steps that would normally require manual browsing.

How It Differs From Other Agents

The Verge notes that Claude Duet is designed specifically for professional workflows, unlike personal assistants such as Instinct that focus on errands like ticket booking. According to TechCrunch, the agent can connect to more than 6,000 connectors and plugins in the Claude Marketplace, and it can make online purchases using Stripe’s Link or Shopify’s Shop Pay when the user authorizes the transaction. While competitors like Meta’s Muse and OpenAI’s Dots also target business users, Claude Duet distinguishes itself by delivering proactive recommendations and by keeping its browser activity hidden from the user, the work happens in the background, and the agent reports back only when it needs approval.

Security and Safety Features

Because the agent may handle sensitive data, Anthropic has added a dedicated classifier and safety‑system prompt. The Verge reports that the classifier is designed to reject requests that involve the strictest tier of information, such as credit‑card numbers. TechCrunch adds that, by default, the agent will turn down requests for credit‑card numbers or government IDs. Before taking any high‑impact action, such as sending a payment or deleting data, the agent messages the user for explicit consent. Anthropic acknowledges that the early version may misinterpret threads or misjudge urgency, so users are advised to stay attentive to the agent’s prompts.

Getting Started and Practical Steps

If you are interested in trying Claude Duet, follow these steps:

• Check eligibility, You need to be a Pro or Max subscriber in the United States. Verify your subscription status in the Claude web portal.
• Join the waitlist, Visit the Claude Duet page on Anthropic’s site and add your email to the waitlist.
• Set up connected apps, Once you receive an invitation, link the applications you use most (email, calendar, CRM, etc.) so the agent can gather data for its briefings.
• Review safety settings, Enable the built‑in classifier and confirm that the agent must ask before any payment or data‑deletion action.
• Start with simple tasks, Begin by assigning low‑risk activities like meeting scheduling or brief report generation to understand how the agent responds.
• Monitor notifications, Pay close attention to the agent’s alerts, especially those that request approval for sensitive actions.
• Provide feedback, As a research preview, Anthropic encourages users to report misreads or false urgencies to help improve the system.

What This Means for Small‑Business Owners

Claude Duet represents another option for businesses that want to offload routine administrative work to an AI system. By automating brief creation, dashboard assembly, and data‑driven forecasts, the agent can free up time for higher‑value activities. However, because the tool is still in preview, owners should treat its outputs as suggestions rather than final decisions and maintain a manual review process for any action involving finances or personal data. The built‑in safety classifier provides a layer of protection, but it is not a substitute for solid internal controls.

Looking Ahead

Anthropic plans to continue refining Claude Duet based on user feedback. Future updates are expected to improve urgency detection, expand connector support, and enhance the transparency of the agent’s background browser work. For businesses that adopt the preview now, early participation offers the chance to shape the feature set and to integrate a tailored AI assistant into daily workflows before broader market adoption. Keeping an eye on Anthropic’s announcements and participating in the feedback loop will be key to getting the most benefit while managing risk.

Overall, Claude Duet adds another tool to the growing ecosystem of AI agents that aim to act on behalf of users rather than simply respond to prompts. For small‑business teams looking to streamline repetitive tasks, the agent’s combination of daily briefings, automated workflows, and safety safeguards makes it worth a trial, provided the organization follows the practical steps above and remains vigilant about approvals and data handling.