Anthropic’s Cowork turns Claude into a truly hands-on digital teammate

Image Credit: Anthropic

Anthropic has unveiled a new feature that brings the power of Claude’s advanced automation tools to everyday users, not just developers or engineers. The tool, called Cowork, is designed to handle real work directly on a user’s computer, making it possible to manage files, organize projects, and complete routine tasks without writing a single line of code.

At its core, Cowork expands on what users can already do with Claude by giving the AI permission to interact with local folders and documents. Once access is granted, Cowork can read files, make edits, or remove content entirely based on user instructions. The idea is to reduce the friction between asking an AI for help and actually getting work done.

One of the standout features of Cowork is its ability to create new projects using existing information. For example, users can point the tool toward a folder full of notes and ask it to generate a structured report. Anthropic says Cowork can also handle practical tasks such as sorting old receipts, building expense-tracking spreadsheets, or cleaning up cluttered download directories that have grown out of control over time.

Cowork also works with Claude’s existing connectors, which means it can pull data from third-party apps to create documents, presentations, and other work files. When paired with the Claude for Chrome extension, the tool can even complete actions that require browser access, blurring the line between an AI assistant and a true digital coworker.

According to Anthropic, Cowork is built to make starting new work with Claude as smooth as possible. Users do not need to repeatedly explain context or manually move outputs into the right format. Instead, they can queue multiple tasks and let Claude work through them at the same time. The experience is meant to feel less like chatting with a bot and more like collaborating with a teammate who understands the job.

That deeper level of access also comes with trade-offs. Anthropic acknowledges that allowing an AI to interact with local files introduces new risks, especially around privacy and data security. To address this, Cowork can only access files and folders that users explicitly approve. Even so, the tool is capable of deleting files if instructed, which is why Anthropic encourages users to be precise when issuing commands.

The company also highlights the potential danger of prompt-injection attacks, which become more serious when an AI has access to sensitive information. While Anthropic says it has implemented safeguards to reduce these risks, it admits that defending against prompt injection remains an active area of research.

For now, Cowork is rolling out as a research preview for Claude Max subscribers on macOS. Users on other plans or operating systems can sign up for a waitlist, with broader availability planned for the future.

Facebook
Twitter
Pinterest
Reddit
Telegram