Unlike a typical chatbot that only responds to text prompts, Anthropic's Cowork acts as an autonomous agent. It can perform tasks, manage files, and execute multi-step plans directly on your computer. This functionality is accessed through the Claude Desktop app for macOS.

Before you can use Cowork, you'll need to install the latest version of the Claude Desktop app. If you already have it, make sure it's updated, as Cowork is introduced as a research preview and may not be on older versions.

Within the desktop app, you'll find different modes of operation. Cowork is often labeled as "Tasks" in the interface. When you select it for the first time, you'll need to initialize it. This is a one-time setup process.

Granting an AI access to your files requires a strong security model. Cowork doesn't run directly on your macOS system. Instead, it operates inside a sandboxed Linux virtual machine (VM). This creates a strict barrier between the agent and your host operating system.

This isolation means Cowork has no visibility or access to your broader file system, applications, or network activity. Its entire world is the Linux VM and the specific folder you authorize.

When you start a task, Cowork will ask for permission to access a specific folder on your local machine. You grant explicit read, write, and delete permissions for that folder only. The agent cannot operate outside this designated workspace.

The Claude Desktop app provides a unified interface for different ways of interacting with the AI. You can easily switch between them depending on your needs. Each mode is tailored for a specific type of work, from simple conversation to complex automation.

Understanding these distinctions is key. Chat is for talking, Code is for programming, and Cowork (Tasks) is for getting work done on your files. The setup you just completed enables this powerful new way of working, transforming Claude from a respondent into a collaborator.