Anthropic has rolled out a new productivity feature called Cowork, expanding the capabilities of its AI assistant beyond chat-based interactions and into direct file management. The feature is now available as a research preview within the Claude macOS app, exclusively for subscribers to Claude Max.
Cowork allows users to grant Claude controlled access to a specific folder on their computer, enabling the assistant to read, edit, and create files directly. This represents a notable shift for Anthropic, moving AI-powered automation out of developer-focused tooling and into workflows designed for a broader audience, including non-technical users.
From chat to delegated work
With Cowork, users can assign multi-step projects to Claude rather than handling tasks through isolated prompts. Examples include organising large folders, extracting structured data from images or documents, and assembling reports from scattered notes. Tasks can be queued, adjusted mid-process, and refined through real-time feedback, offering a more continuous and collaborative workflow than traditional chatbot interactions.
The feature also supports skills and connectors that allow Claude to integrate external data sources or interact with browser-based workflows. Anthropic says this approach is meant to resemble delegating work to a digital collaborator rather than issuing one-off commands.
Platform limits and expansion plans
At launch, Cowork is limited to macOS users in the United States. Anthropic has confirmed plans to expand support to Windows and introduce syncing capabilities in future updates, though no specific timeline has been announced.
Under the hood, Cowork is built on the same agent foundations as Claude Code, Anthropic’s coding-focused assistant. However, the company says the consumer-facing version includes additional safety layers and user controls tailored to everyday file operations.
Emphasis on control and safety
Anthropic has placed a strong emphasis on permissions and transparency. Users explicitly choose which folders or connectors Claude can access, and the assistant must request approval before taking sensitive actions such as deleting files. These safeguards are designed to minimise the risk of unintended changes and address concerns related to prompt injection or overreach.
Early feedback from developers and productivity communities suggests Cowork could enable a new class of workplace automation, particularly for professionals who manage large volumes of documents or repetitive organisational tasks. At the same time, Anthropic is positioning the feature as an experiment, gathering user input as it refines security, usability, and scope.
As AI assistants increasingly move beyond conversation and into direct interaction with users’ files and systems, Cowork signals Anthropic’s ambition to redefine how people collaborate with AI in day-to-day work—turning Claude from a helpful responder into an active participant in digital workflows.
