Agent mode in ChatGPT was a tool that let ChatGPT do multi-step tasks for you: browse websites, click buttons, fill in forms and use connected apps, instead of only answering in chat.
One important update first: OpenAI’s own ChatGPT agent help page now says “ChatGPT agent is no longer available” and points users to ChatGPT Work for longer, multi-step tasks. So this guide explains what agent mode was, what replaced it, and how to use the replacement today. Checked against OpenAI’s help pages in November 2026.
What was agent mode in ChatGPT?
According to OpenAI’s help page, agent mode worked like this:
- You picked it from the tools menu or typed /agent, then described a task.
- It used a virtual browser, seeing pages through screenshots, so it could click, scroll and fill in forms.
- It paused to ask for clarification or confirmation when needed, and you could interrupt it mid-task.
- If a site needed a login, it asked you to take over the browser so you typed the password yourself.
- It was only on paid plans. The help page lists monthly limits such as 40 agent messages on Plus and 400 on Pro.
- Finished tasks could be set to repeat daily, weekly or monthly.
OpenAI also says the older Operator tool was folded into agent mode.
What replaced it: ChatGPT Work
OpenAI’s ChatGPT Work and Codex help page describes Work as “an agent designed for longer, multi-step work and finished deliverables.” Chat is for quick questions. Work is for tasks like research, analysis, documents, spreadsheets, presentations and reports.
Per OpenAI, Work is available on desktop for all plans and on web and mobile for Plus, Pro, Business, Enterprise and Edu. Check ChatGPT’s pricing page for what your plan includes.
How to use ChatGPT Work (step by step)
- Open ChatGPT on the web, mobile app or desktop app and sign in.
- Select Work instead of Chat.
- Start a new chat, or open a project if the task is part of ongoing work.
- Add any files or context it should use.
- Describe the outcome you want, the sources to use, any limits, and when it should stop for your approval.
- Follow its progress, answer questions, and approve important actions.
- Review the result and ask for changes in the same chat.
Tips for better results
OpenAI’s getting started guide gives this advice, in plain words:
- Say what the finished result should look like.
- Set clear limits, like “use only these sources” or “stop before sending anything.”
- Use Work for big multi-step tasks. Use Chat for quick questions and short rewrites.
- Longer tasks can use more of your plan’s usage, so check your limits before a big job.
Good tasks to try
- Research and compare three tools for your side hustle, with links.
- Turn your notes into a one-page plan or a simple spreadsheet.
- Draft a weekly content calendar from a list of topics.
- Set up a recurring summary of news in your niche.
Safety: be careful what you connect
OpenAI warns that when an agent can sign into sites or use apps, it can reach sensitive data, and there are risks such as “prompt injection,” where a web page tries to trick the agent. Its advice includes:
- Only connect the apps a task really needs.
- Don’t type passwords or private info into the chat.
- Avoid vague prompts like “check my email and handle everything.”
- Stop the task if anything looks wrong.
Where to go next
For more ways to save time with AI, see our AI and productivity hub and our guide to the best AI tools for writing. If you are using AI to build extra income, start with our Start Here guide.
FAQ
Is agent mode still available in ChatGPT?
No. OpenAI’s help page says ChatGPT agent is no longer available and recommends ChatGPT Work instead.
Was agent mode free?
No. OpenAI said agent mode was only on paid plans.
What is the difference between ChatGPT Chat and Work?
Chat is for answers, brainstorming and short drafts. Work is for longer tasks that end in a finished file or result.
Are there free AI agents?
ChatGPT Work is available on desktop on all plans, including Free, according to OpenAI. Limits apply, so check the pricing page.

