OpenAI’s ChatGPT Agent is now available on the ChatGPT Mac app, so you can automate task execution directly on your desktop. Previously limited to web users, the Mac implementation lets users trigger the agent using either the /agent command or by tapping a new Agent button at the bottom of the interface. Once activated, the agent spins up a visual workspace and executes tasks step by step, while displaying a reasoning timeline so you can track progress or intervene at any time.
This isn’t just a simple assistant—it combines OpenAI’s Deep Research for parsing and interpreting web content with Operator, which automates actions such as navigating websites, clicking buttons, and interacting with external platforms like Google Drive or GitHub. The agent lets you hand off complex workflows—think wedding planning, multi-site product research, or automated slide creation—while you step away and check back in later. These autopilot-style operations mimic what a junior intern might do, but faster and more accurately.
In one demo scenario, Agent browsed clothing sites, searched hotel prices, scheduled events, created a spreadsheet of options, and plotted routes—all without human input. In another, it pulled data from Google Drive, generated a full presentation with charts, and even started an order process but paused before final checkout to ensure control remained with the user.
Access is currently limited to Pro, Plus, and Team subscribers—Pro users get up to 400 agent queries per month, while Plus and Team members are capped at 40. Enterprise and educational access is expected by the end of July.
Bottom line: the Mac app upgrade brings powerful AI autonomy to desktop users, offering a hands‑off productivity boost for tasks that span browsing, coding, planning, and content creation. It’s a leap forward for AI productivity tools—but mastering it means figuring out which workflows are best suited to the Agent mode.