ChatGPT Agent

ChatGPT Agent (also referred to as "agent mode") is OpenAI's autonomous web-task agent integrated directly into ChatGPT. It evolved from the standalone Operator product launched in January 2025, which was sunset in mid-2025 and folded into ChatGPT itself in July 2025 -- the operator.chatgpt.com URL no longer resolves. Users invoke it by selecting agent mode from the tools menu or typing /agent in the composer. Powered by OpenAI's Computer-Using Agent (CUA) model -- a vision + reasoning system that "sees" web pages via screenshots and "acts" through simulated mouse and keyboard -- ChatGPT Agent can navigate websites, fill out forms, edit spreadsheets, work with uploaded files, connect to third-party data sources (email, document repositories), run code via the code interpreter, and execute terminal commands. Tasks typically complete in 5-30 minutes. Critically, agent mode pauses and hands control back to the user for sensitive steps: logins, payment information, CAPTCHAs, and high-impact actions like submitting orders or sending emails. Watch mode requires close supervision on financial or email sites. Recurring tasks can be scheduled daily/weekly/monthly. Pricing is included in paid ChatGPT plans with monthly message limits: Plus ($20/mo) gets 40 agent messages/month; Pro ($200/mo) gets 400; Business ($25/user/mo) and Enterprise get 40/user with flexible-pricing options at 30 credits per message. No free tier. Available on web, iOS, Android, and desktop. Enterprise workspace owners get RBAC, website blocking, allowlisting, EU data residency, and Compliance API logging. Best fit: knowledge workers running repetitive web tasks (form filling, data extraction, scheduled reports), revenue ops teams automating multi-step workflows, and enterprises needing autonomous web execution with human checkpoints. Not appropriate for high-volume scraping (use Browse AI or Skyvern), banking transactions (refused by design), or teams requiring sub-minute response times. Not affiliated.