ChatGPT Agent Mode Is Gone: What to Use Instead
ChatGPT agent mode has been removed. OpenAI’s help centre now opens with it plainly: “ChatGPT agent is no longer available. Use ChatGPT Work for longer, multi-step tasks and finished deliverables. For supported browser workflows, see Using cloud browser in ChatGPT.”
If you came here because the option vanished from your tools menu, that is why. You are not looking at a bug, a rollout, or a plan downgrade.
Two things make this more confusing than a normal deprecation, and both are worth knowing before you try to rebuild anything.
OpenAI’s own documentation still contradicts itself
The ChatGPT agent help article carries that “no longer available” notice at the top, and then continues, unchanged, to explain how to use the feature: “To start using agent mode, select it from the tools menu or type /agent in the composer.” It still lists availability on “Pro, Plus, Business, Enterprise, and Edu plans” and still publishes the old monthly limits of 40 messages on Plus and 400 on Pro.
The deprecation notice was bolted onto the old article rather than replacing it. So if you followed the instructions on the official page and found nothing in your composer, the page was wrong, not you.
ChatGPT Work is not a like-for-like replacement
This is the part that matters if you had a working setup. Work handles research, public web browsing, files, and document creation. What it does not do is the thing agent mode was uniquely good at: driving a real browser through a site you are logged into, and handing control back to you mid-task.
Agent mode spawned a virtual computer, opened a real browser, visually interpreted rendered pages, and clicked through interfaces. It paused at logins and payments and gave you the browser to finish that step by hand, then carried on. That handoff is what made logged-in workflows possible at all: product configurators, shipping options, account settings, anything behind a sign-in.
Users making this point in OpenAI’s developer forum on August 8, 2026 note that Deep Research, Codex, Work, and computer use in ChatGPT each cover part of the old surface and none reproduces that one. The same thread notes there was no advance notice, no deprecation period, and no migration guide.
For browser-based work, OpenAI’s pointer is the cloud browser in ChatGPT rather than Work. Check whether your specific workflow is supported there before assuming Work covers it.
If you had a scheduled task using agent mode
This is the most common way people discovered the removal: a recurring task that had been running silently for months stopped producing anything.
Agent mode was invocable from scheduled tasks, and OpenAI’s own limits documentation confirms it counted that way, noting that “each unique agent invocation will count against the monthly message limit. This includes agent requests that are part of scheduled tasks.”
There is no automated migration. To audit what you have:
- Go to
chatgpt.com/schedules. - Review every recurring task and identify which ones invoked agent mode.
- Check the last few runs of each. A task that silently returns nothing useful is the signature.
- Rebuild the ones that matter, deciding per task whether Work, the cloud browser, or Codex fits.
- Delete or pause the ones that cannot be rebuilt, so they stop consuming schedule slots.
Step 3 is the one people skip. A broken scheduled task does not announce itself; it just quietly stops being useful, which is why some of these ran dead for weeks.
What agent mode actually did, for reference
Useful if you are trying to work out which replacement fits, or if you are reading old documentation that still describes it as current.
Agent mode launched in July 2025 and absorbed two earlier products: Operator, the browsing agent, and Deep Research, the long-form research agent. Neither exists separately either, which is why searches for “ChatGPT Operator” have been landing on agent mode documentation. It could:
- browse the web, navigating and clicking like a person at a keyboard;
- open and reorganize documents you uploaded;
- connect to email and document sources for context;
- complete and submit web forms;
- build and modify spreadsheets as part of a task.
Tasks typically finished in 5 to 30 minutes. It was paid-only, monthly-capped, and supervised, which made it a project runner rather than an assistant. Calendar work was its most consistently reported weak point, where time zones and meeting conflicts need judgment a browsing agent does not have.
The structural gap none of the replacements close
Worth being clear that this is not what was taken away in August. It was never there.
Give Claude's Cowork tasks a time and they will run. Give them something that just happened and they will not. ChatGPT goes further, but only as far as Gmail, Slack and GitHub, only on web and mobile, and never alongside a schedule on the same task. A lead moving stage, a form landing, an invoice clearing: none of it starts either one. Carly starts on all of it.
Agent mode was a session you launched. So is Work. Neither watches anything and starts on its own when something happens, which is the difference between a tool you operate and an assistant that handles a recurring job.
If what you actually wanted automated is the inbox-and-calendar grind rather than a bounded research project, that gap is where Carly sits. It fires on events: when mail arrives, when an invite lands, on a schedule. It works Gmail and Outlook, drafts and sends replies with attachments, files and labels, updates a CRM, and each agent gets its own email address so it can correspond as a colleague rather than hand work back at a login wall. Setup is a conversation rather than a config screen. Carly offers free Zapier-style workflows; AI agents start at $35/month.
Picking a replacement
| What you were doing in agent mode | Where it goes now |
|---|---|
| Multi-step research, then a deliverable | ChatGPT Work |
| Browsing public sites and gathering data | ChatGPT Work |
| Driving a site you have to log into | Cloud browser in ChatGPT, if supported |
| Handing the browser back to finish a login | No direct equivalent |
| Filling repetitive web forms | Cloud browser, or a scripted approach |
| Spreadsheet building and cleanup | ChatGPT Work |
| Code-shaped or scripted automation | Codex |
| A recurring job that should start on its own | Neither; needs a trigger-based agent |
Frequently Asked Questions
Is ChatGPT agent mode gone?
Yes. OpenAI’s help centre states that ChatGPT agent is no longer available and directs users to ChatGPT Work for multi-step tasks and to the cloud browser for browser workflows. The notice appeared in early August 2026.
Why does OpenAI’s help page still explain how to use it?
Because the deprecation notice was added to the top of the existing article without removing the body. The instructions, the plan availability list, and the old monthly message limits are all still published beneath a notice saying the feature is unavailable.
Is ChatGPT Work the same as agent mode?
No. Work covers research, public browsing, files, and document creation, but not driving a logged-in browser session or handing control back to you mid-task. If your workflow depended on signing into a site, test the cloud browser rather than assuming Work replaces it.
What happened to ChatGPT Operator?
Operator was folded into agent mode in 2025, along with Deep Research, and agent mode has now been removed. There is no separate Operator product to return to.
My scheduled task stopped working. What do I do?
Open chatgpt.com/schedules, find the tasks that invoked agent mode, and check their recent runs. There is no automated migration, so each one has to be rebuilt against Work, the cloud browser, or Codex, or paused if no replacement fits.
Will agent mode come back?
OpenAI has not said it will. The forum thread asking for restoration or a functional equivalent has no staff response, and the help centre presents the removal as final rather than temporary.
More: What is ChatGPT Work · ChatGPT Work review · ChatGPT scheduled tasks · What are AI agents · Best AI agents for productivity · Can ChatGPT schedule meetings?
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
