OpenAI just launched ChatGPT Agent, a significant leap forward that merges the capabilities of its previous tools—Operator and Deep Research—into a single, more powerful assistant. Announced during a livestream featuring Sam Altman and the OpenAI team, this new mode allows the AI to carry out complex, multi-step tasks by using its own virtual computer.
In the demonstration, the Agent handled real-world scenarios with ease: it reviewed calendars, identified free time slots, booked restaurant reservations, planned meals and ordered grocery ingredients, generated research reports on competitors, filled online forms, and even created slide decks based on live data. Despite occasional delays—tasks can sometimes take 10–30 minutes—the payoff is dramatic time savings compared to manual execution.
Crucially, OpenAI emphasised safety and user control: the agent always requests explicit approval before taking irreversible actions, includes a “watch mode” for sensitive tasks, and is built to resist malicious prompt injection through hidden monitoring systems.
Access is rolling out immediately to Pro, Plus, and Team subscribers in the US, with Pro users receiving up to 400 agent calls per month, and others getting 40. Enterprise and Education tiers are expected later in the summer, while EEA and Switzerland are not supported yet.
This release marks a clear push by OpenAI to define the AI agent space by delivering a unified, action‑oriented assistant that can think, act, and collaborate like a digital co‑worker.
In summary, ChatGPT Agent is not just a chatbot—it’s a virtual employee equipped with browsing tools, file and code execution, research synthesis, and task automation. It operates transparently, keeps you in the loop, and accelerates workflows across personal and professional tasks. Whether it’s planning a date night, compiling financial slideshows, automating weekly reports, or managing bookings, this new agentic mode delivers — albeit sometimes at the cost of waiting a few minutes.
With Agent, OpenAI makes good on its vision of AI assistant capabilities, and as the rollout continues, expect access to expand and for this to become deeply integrated into everyday productivity.
