OpenAI has launched a new AI tool called ChatGPT Agent, which can perform complex tasks on its own using a virtual computer. This feature is now available to paid ChatGPT users, including Plus, Team, and Pro subscribers.
The ChatGPT Agent brings together two earlier tools—Operator and Deep Research—into one powerful system. It can browse websites, click buttons, search for specific information, write code, and even use third-party apps like Gmail and GitHub. OpenAI showed off this new feature in a recent live stream.
What makes the ChatGPT Agent special is that it can work independently. For example, it can:
- Turn a screenshot into a PowerPoint slide with editable elements
- Add new data to spreadsheets without breaking the format
- Book meeting rooms or plan full events
- Research local tax laws and create early retirement plans
- Plan trips, parties, or even find doctors and book appointments
The Agent uses different tools to do these tasks, including:
- A visual web browser to view and click on websites
- A text browser for quick searches
- A coding terminal for programming tasks
- API access to connect with other software
The Agent works inside the ChatGPT app on web and desktop. Users can chat with it and then ask it to take action—all in one place. When the agent starts a task, users can watch what it’s doing in a small window inside the chat.
Right now, Pro users can send up to 400 requests to the agent each month. Other paid users will get access soon, with a monthly limit of 40 messages. The Agent will also be added to Enterprise and Education plans soon.
OpenAI said that this feature is not yet available in the European Economic Area and Switzerland, but they are working to add support.
With this update, OpenAI will also shut down the older Operator tool, as ChatGPT Agent now includes its abilities and more.
Also Read: Meta Fixes AI Bug That Could Expose Private Chats
