OpenAI has announced a new agentic system for ChatGPT which can now “do work for you using its own computer,” handling various complex tasks for a user.
The new ‘agent mode’ can handle requests like analyzing a user’s calendar, creating slideshows, submitting expenses, or generating code.
It combines several capabilities of OpenAI’s earlier agentic tools, such as Operator’s ability to interact with websites, deep research’s skill in synthesizing information, and ChatGPT’s intelligence and conversational fluency.
ChatGPT relies on its own virtual computer and shifts between reasoning and action to complete complex tasks.
ChatGPT can now do work for you using its own computer.
undefined OpenAI (@OpenAI) July 17, 2025
Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths. pic.twitter.com/7uN2Nc6nBQ
OpenAI says that you, as the user, always remain in control of the process, able to interrupt or over-take the task at any point.
Pro, Plus, and Team subscribers can activate the mode through the tools dropdown by selecting ‘agent mode’. According to OpenAI, you can simply describe the task, and ChatGPT will perform it, adding an on-screen narration of what it's doing. You are free to interrupt it whenever needed to ensure the chatbot does exactly what you are intending it to.
ChatGPT agent is equipped with connectors, allowing you to connect apps like Gmail and Github. This enables ChatGPT to find relevant information directly and use it in response to your prompts.
You can also log in on any website to expand the chatbot’s task execution scope. Some tasks can be scheduled to recur automatically in case you need weekly reports or regular updates.
“For instance, it can gather information about your calendar through an API, efficiently reason over large amounts of text using the text-based browser, while also having the ability to interact visually with websites designed primarily for humans,” OpenAI explains.
Among the examples of what ChatGPT agent can take care of are requests like “look at my calendar and brief me on upcoming client meetings based on recent news,” “plan and buy ingredients to make Japanese breakfast for four,” and “analyze three competitors and create a slide deck.”
The model’s capabilities are reflected in its state-of-the-art (SOTA) performance in several areas, according to OpenAI. For instance, On Humanity’s Last Exam, which evaluates a model’s performance across a broad range of subjects on expert-level questions, ChatGPT agent scored a new pass@1 SOTA at 41.6%, which could be pushed to 44.4%.
In a blog post, OpenAI admitted that the new functionality introduces novel risks, especially because ChatGPT can now work directly with user data. Although the company strengthened the safeguards and added mitigations that significantly reduce risk, OpenAI admits that the agent’s wider range of tools and functionality increases its overall risk profile.
“We have built a lot of safeguards and warnings into it, and broader mitigations than we’ve ever developed before from robust training to system safeguards to user controls, but we can’t anticipate everything,” said OpenAI CEO Sam Atman in a post on X.
“I would explain this to my own family as cutting edge and experimental; a chance to try the future, but not something I’d yet use for high-stakes uses or with a lot of personal information until we have a chance to study and improve it in the wild.”
Today we launched a new product called ChatGPT Agent.
undefined Sam Altman (@sama) July 17, 2025
Agent represents a new level of capability for AI systems and can accomplish some remarkable, complex tasks for you using its own computer. It combines the spirit of Deep Research and Operator, but is more powerful than that…
Your email address will not be published. Required fields are markedmarked