Anthropic Launches Claude Duet, a Proactive AI Assistant

By Suad Seferi · 22 October 2026 at 18:22

Anthropic introduced Claude Duet on October 21, 2026, a new AI assistant designed to monitor connected applications, identify tasks that need attention and complete work without requiring users to remain in an active conversation. Unlike conventional AI chatbots that wait for instructions, Claude Duet can initiate contact, prepare daily briefings and continue working while users are away from their computers. The product is entering a limited research preview for Claude Pro and Max subscribers in the United States, with access managed through a waitlist. It is available as a separate application for iOS, Android and the web. An AI assistant that comes to you Claude Duet is powered by Anthropic's Claude Opus 5.5 model and is designed primarily for professionals and small business owners. Each morning, the assistant prepares a briefing based on information from connected applications, including email, calendars and workplace communication tools. It identifies upcoming commitments, unfinished tasks and information that may require a decision. Throughout the day, Duet checks connected accounts every few hours and contacts users when it identifies something important. For example, an assistant connected to a business email account and calendar could notice an upcoming client meeting, collect relevant information and prepare a briefing before the meeting takes place. It could also identify unpaid invoices, draft payment reminders or prepare a follow-up message after a customer interaction. Users can assign more substantial tasks, including creating financial forecasts, building dashboards, developing prototypes and generating web applications. These capabilities move Claude beyond answering questions and generating content. The assistant is intended to take responsibility for completing parts of a workflow. How Claude Duet works Duet operates through a dedicated virtual computer in the cloud, equipped with its own browser. This allows the assistant to continue working even when the user's laptop or smartphone is switched off. Rather than requiring users to manage multiple conversations, Duet maintains a continuing interaction that incorporates information from their existing Claude account, including memory, skills and connected applications. Anthropic says the assistant can work with more than 6,000 connectors and plugins available through the Claude Marketplace. These include services such as Gmail, Slack, QuickBooks, Stripe, HubSpot, Canva and Shopify. A business owner could theoretically ask Duet to review outstanding invoices, prepare reminders, update customer records and compile a report using information from several connected services. However, the availability of integrations does not guarantee that every workflow can be completed reliably or without additional configuration. What happens when AI starts making decisions? Giving an AI assistant access to business applications introduces security and accountability questions. An agent that can read correspondence, access financial records and prepare transactions has considerably more operational responsibility than a chatbot generating text. Anthropic says Claude Duet requires explicit approval before performing sensitive actions, including sending messages, making payments, signing documents and deleting files. Certain operations, such as trading securities or transferring money between accounts, remain prohibited even with user approval. The company also says Duet cannot directly access passwords or authentication tokens and uses additional safeguards intended to prevent sensitive information from being disclosed. These protections include a dedicated security classifier, defenses against malicious instructions embedded in external content and penetration testing of the system's infrastructure. Nevertheless, such safeguards do not eliminate the risks associated with autonomous agents. An assistant could misunderstand instructions, incorrectly prioritize information or produce an inaccurate document. Malicious content encountered through connected applications could also attempt to influence its behavior. Anthropic acknowledges that the preview version can misinterpret conversations and make mistakes when deciding which issues require immediate attention. For businesses, this makes human review particularly important when AI-generated work involves customers, financial transactions or confidential information. Availability and limitations Claude Duet is not yet available to subscribers in North Macedonia or other Balkan markets. The initial research preview is restricted to US-based Pro and Max subscribers. Team and Enterprise plans are not included. Anthropic has not announced a timetable for broader international availability. During the preview, Duet shares the usage limits of the user's existing Claude subscription. Its background checks consume those limits, meaning that additional activity by the assistant can reduce the capacity available for other Claude tasks. Duet also does not continuously inspect connected applications in real time. Its periodic checks mean that important information may be detected some time after it arrives. The company describes the product as an early version, with some capabilities still under development. What this means for businesses in the Balkans Although Claude Duet is currently unavailable in the region, its introduction illustrates how AI tools are moving toward more active participation in business operations. For smaller companies, the potential appeal is straightforward. Tasks such as following up on invoices, preparing meeting materials, reviewing communications and compiling reports consume working hours even when they do not require specialist expertise. Delegating parts of that work to an AI assistant could reduce administrative effort. But practical adoption requires more than connecting an account. Businesses would need to determine which information an AI agent may access, which tas…