Anthropic has introduced Cowork, a new research preview tool designed to make its AI capabilities more accessible to non-technical users. Built into the Claude Desktop app for macOS, Cowork allows users to grant Claude access to a specific folder on their computer, enabling the AI to read, edit, or create files within that space using simple chat instructions. This approach offers a user-friendly alternative to the more complex command-line interface of Claude Code, which was launched in November 2024 and has since been used for a wide range of non-coding tasks. Cowork is based on the same underlying model and technology as Claude Code, including the Claude Agent SDK, but it simplifies the process by eliminating the need for technical setup like virtual environments or command-line tools. Users can now perform complex, multi-step tasks—such as organizing a folder of receipt images into an expense report, re-arranging media files, or compiling a report from scattered notes—without writing code or managing file formats manually. The tool operates with a high degree of autonomy, creating a plan, executing actions, and keeping users informed throughout, much like a collaborative coworker. Unlike a standard chat conversation, Cowork enables continuous, goal-driven workflows. Users can queue up multiple tasks and let Claude work on them in parallel, reducing the back-and-forth typically required. The experience is more about delegation than dialogue, allowing users to focus on higher-level direction while Claude handles the execution. Security and control remain central to the design. Users can choose exactly which folders and external connectors (like email or cloud services) Claude can access, and the AI will not act on any file outside of those permissions. Before taking major actions—such as deleting or overwriting files—Claude will prompt for confirmation, giving users the ability to intervene. However, Anthropic cautions that clear, unambiguous instructions are essential, as misinterpretation can still lead to unintended consequences, including data loss. The company also highlights the risk of prompt injection attacks, where malicious content could influence Claude’s behavior, even if the AI has built-in defenses. While these risks are not new, they become more significant as tools like Cowork gain greater real-world control, making them a key focus for ongoing safety research. Cowork is currently available only to Claude Max subscribers on macOS, with a waitlist for users on other plans. Anthropic is releasing it as a research preview to gather user feedback and identify new use cases, with plans to improve the tool by adding cross-device sync, support for Windows, and enhanced safety features. The tool reflects a growing trend: users are increasingly applying AI agents to non-technical, real-world tasks. By making these capabilities more intuitive, Anthropic aims to expand the reach of its AI beyond developers and into everyday productivity, from personal organization to content creation. As the company continues to iterate, Cowork could become a central part of how people interact with AI in their daily work, blending automation, agency, and user control in a way that feels both powerful and approachable.