6,000 Connectors: The Scale Behind Anthropic's New Business Agent

For professionals and small business owners, the landscape of AI assistance is shifting from simple chatbots to autonomous workers. As of late 2026, Anthropic is introducing Claude Duet, a new application designed to handle complex operational tasks rather than just answering questions. This launch marks a distinct pivot in how the company approaches the crowded market of AI assistants, positioning its tool not as a general-purpose personal aide, but as a specialized engine for workflow automation.

The primary distinction of this new agent is its focus on professional utility over personal convenience. Anthropic explicitly notes that Duet is not intended for handling casual errands, such as booking tickets or managing personal schedules. Unlike some competitors that allow users to watch the agent’s browser activity in real-time, Duet operates largely behind the scenes. It utilizes its own virtual machine to complete tasks in the background, only surfacing to the user when specific approvals or inputs are required. This design choice suggests a preference for efficiency and minimal interruption for working professionals.

A significant technical advantage lies in the agent’s connectivity. Built on the Claude Opus 5.5 model, Duet can interface with more than 6,000 connectors and plug-ins available in the Claude Marketplace. This extensive integration capability allows the agent to build daily briefings, generate data dashboards, and suggest actions based on user-specific data. By proactively monitoring ongoing tasks, the system delivers recommendations without requiring constant manual prompting, effectively acting as a digital manager for routine operations.

Financial transactions represent a key area of functionality, though with strict boundaries. The agent can make online purchases on behalf of users through integrations with Stripe’s Link digital wallet or Shopify’s Shop Pay. However, Anthropic has implemented safeguards to limit risk. Sensitive steps, particularly those involving payments, require explicit user approval. Furthermore, the agent is restricted from trading securities or moving money between bank accounts, ensuring that its financial autonomy remains within a defined, safer perimeter.

Security and privacy remain central to the product’s architecture. To address concerns about data leakage, Anthropic has released a new classifier that scrutinizes each message. By default, the agent is configured to decline requests involving sensitive data, such as credit card numbers or government identification, unless explicitly permitted. This layer of protection is critical for a tool that processes professional data and executes tasks autonomously.

The market context for this launch is one of rapid evolution. The industry is moving beyond passive AI assistants toward active agents that work on a user’s behalf. While rivals like Meta and OpenAI are also expanding into business-focused AI, Duet’s specific targeting of small businesses and professionals sets it apart from purely personal assistants. This strategic positioning aims to capture a segment of users who require robust, integrated tools for their daily work lives.

Currently, Claude Duet is available in a research preview for Pro and Max customers on iOS, Android, and the web via Claude.ai. Access is limited to those who join the waitlist, indicating that the company is managing rollout carefully. As the distinction between assistants and agents blurs, the success of Duet will likely depend on how well it balances autonomous capability with the trust required to handle professional tasks and financial data.