On Monday, Anthropic unveiled "CoWork," a new AI feature that brings the power of its successful tool "Claude Code" to non-technical users. Interestingly, the development team was able to create this feature in less than two weeks with the help of the same Claude Code AI.
CoWork marks a milestone in the competition to provide practical AI assistants for general users and positions Anthropic not only against OpenAI and Google but also in competition with Microsoft and its smart productivity assistants.
This product allows users to designate a folder on their computer without any programming knowledge so that the AI can read, edit, or even create new files. For example, CoWork can be asked to turn scattered receipts into a structured expense report or organize and rename a messy downloads folder.
CoWork utilizes an architecture based on the "agentic loop," meaning the AI does not just generate text but also plans, multitasks, reviews results, and asks questions if needed. This structure makes the interaction with CoWork feel more like delegating work to a colleague rather than chatting with a robot.
Another interesting point is the astonishing speed of CoWork's development; a large part of it was written by the Claude Code tool itself, demonstrating a feedback loop where AI assists in its own development, accelerating the creation of AI products.
CoWork also has the ability to connect to services like Asana, Notion, and PayPal, and can access the web through the Claude in Chrome browser extension to perform tasks like filling out forms. Additionally, a set of "skills" has been designed for CoWork to enhance its capabilities in document and presentation creation.
Despite its many features, Anthropic has warned users that CoWork may delete files or take risky actions if given incorrect commands and emphasized that security in using this assistant is still evolving.
Currently, CoWork is available only on macOS for premium Claude (Max) subscribers and is planned to be released for Windows and other systems in the future. Anthropic has stated its goal as providing a friendly and trustworthy AI experience for regular users.