Use Your Computer and Browser

**TL;DR: ChatGPT Work introduces "Computer Use" and "Browser Use" features, allowing the AI to operate desktop applications and browsers on behalf of users, while users retain final approval rights.** ## Overview of ChatGPT Work ChatGPT Work can utilize tools configured by the user for plugins. However, some tools lack plugins or exist within desktop applications and logged-in websites. To address this, OpenAI has launched the "Computer Use" and "Browser Use" features. These are available exclusively in the Work tab of the ChatGPT desktop app—not in the standard version of ChatGPT, web version, or mobile app. ## "Computer Use" Feature "Computer Use" enables ChatGPT to interact with desktop applications just as a user would, including clicking, typing, and switching between different apps and windows. This feature can run in the background, allowing users to continue using their computer. ### How It Works and Examples Before accessing an application, ChatGPT requests permission, and users can choose to allow it once or remember the selection. For example, a user might say via voice input: "Play my Daily Mix playlist on Spotify." Despite lacking a specific Spotify integration, ChatGPT uses the installed desktop app to execute the task. On a Mac, it can process in the background, and users can observe, take over, or stop the task at any time. ### Screenshot Feature Users can press Command+Command to capture an "app screenshot," providing ChatGPT with the context of the current application or website. This helps the AI understand what it's seeing without requiring detailed explanations from the user. ## "Browser Use" Feature "Browser Use" allows ChatGPT to work within the browser the user is already using, including logged-in websites. It handles tasks that typically require opening tabs and filling out forms. ### How It Works and Examples For instance, a user might request: "Submit time-off requests for September 14 and 15 on Workday, check my calendar for conflicts, and draft an out-of-office event for the same dates." ChatGPT uses the browser to access Workday (since no plugin is available), then utilizes the Google Calendar plugin to check the schedule and prepare the event. The task pauses upon completion, awaiting user review. ## Permissions and User Control ChatGPT requests permissions before taking action, allowing users fine-grained control. For actual changes that affect calendars, teams, or other systems—such as submitting requests or creating events—users prefer to retain the final decision. Therefore, ChatGPT first shows a draft and pauses. ## Best Practices and Decision Rules Users share a simple rule: use plugins when directly connected tools are available; choose "Computer Use" when the work involves desktop applications; and opt for "Browser Use" when the work is within logged-in browsers. The core of ChatGPT Work is to handle tedious steps while users review critical actions. Source: Use Your Computer and Browser (https://www.youtube.com/watch?v=981SivztzOc)