For years, AI has been something you talk to in a chat box – a disembodied voice that can write essays or debug code but remains trapped behind a browser window. With the release of Cowork, Anthropic is breaking that fourth wall. Announced this week as a research preview, Cowork is a new agentic feature that lets Claude reach out of the chat interface and directly into the folders on your computer.
Think of it as “Claude Code” for everyone else. While Anthropic’s coding tools have long allowed developers to automate complex technical workflows, Cowork brings that same autonomy to general knowledge work. Instead of pasting text back and forth, you simply give Claude access to a specific folder on your Mac and tell it what to do.
Also read: Elon Musk says Apple-Google AI deal is bad for all of us, here’s why
The core promise of Cowork is agency. Unlike a standard chatbot that waits for your next prompt, Cowork acts like a digital colleague. You might point it at a messy “Downloads” folder and ask it to organize everything by date and file type. Or, you could dump a pile of receipt screenshots into a folder and ask Claude to “create an expense report in Excel.”
Because it has direct read-and-write access to the files in that folder, Claude can open the images, extract the data, and generate a new spreadsheet file right next to them, all without you lifting a finger. It breaks these requests down into a plan, executes them step-by-step, and can even handle multiple tasks in parallel.
Also read: Sam Altman’s problem: Apple’s Gemini pick for Siri disempowers ChatGPT
Cowork isn’t limited to just tidying up your desktop. It integrates with “Claude in Chrome,” meaning it can combine local file access with web research. You could ask it to research a topic online and draft a report in a Word document saved directly to your project folder. It also supports external “connectors,” allowing it to pull in data from other business tools to enrich the documents it creates.
Giving an AI permission to delete or rewrite your files sounds risky and Anthropic knows it. To keep things safe, Cowork operates in a “sandbox.” It can only see and touch the specific folders you explicitly share with it. Furthermore, it is designed as a “human-in-the-loop” system; for any significant or potentially destructive action (like deleting a batch of files), Claude will pause and ask for your confirmation before proceeding.
However, the company warns that risks like “prompt injection” – where malicious web content tricks the AI into misbehaving – are still a concern. Users are advised to supervise the agent, especially during this early preview phase.
Currently, Cowork is available exclusively to Claude Max subscribers using the macOS desktop app. While Windows support is planned, for now, this powerful glimpse into the future of agentic work is a Mac-only experience.
Also read: Anthropic unveils medical AI suite to rival ChatGPT Health: Key features explained
