Operator
Operator was OpenAI's browser agent, a research preview for Pro users in the US launched on Jan 23, 2025 at operator.chatgpt.com. On Jul 17, 2025 it was folded into ChatGPT as agent mode, and the Operator website is no longer accessible.
Operator no longer exists as its own product: on Jul 17, 2025 OpenAI "fully integrated" it into ChatGPT as ChatGPT agent, and the standalone site, operator.chatgpt.com, was shut down; OpenAI's help article now says "The Operator website is no longer accessible" (OpenAI Help Center). Operator, launched on Jan 23, 2025, was a research preview of "an agent that can go to the web to perform tasks for you", using its own browser to type, click and scroll. To do the same today, choose agent mode from the tools menu in ChatGPT or type /agent; see Agent Mode.
Quick facts
| Status | Merged into ChatGPT agent Jul 17, 2025; site no longer accessible |
| Launched | Jan 23, 2025, research preview |
| Who could use it | Pro users in the US, at operator.chatgpt.com |
| Model | Computer-Using Agent (CUA), GPT-4o vision plus reasoning trained with reinforcement learning |
| Replaced by | Agent mode in ChatGPT |
What did Operator do?
Operator ran in a remote browser of its own. It looked at web pages through screenshots and acted with every action a mouse and keyboard allow, so it could handle repetitive browser jobs such as filling out forms, ordering groceries and even making memes (OpenAI). A user described the task and could take control of the browser at any point, for example to sign in. OpenAI said its model set new state-of-the-art results on the WebArena and WebVoyager browser benchmarks at the time.
Operator was powered by the Computer-Using Agent (CUA), which combined GPT-4o's vision with advanced reasoning trained through reinforcement learning to work graphical interfaces: the buttons, menus and text fields people see on screen (OpenAI). In the OpenAI API, the computer-use-preview model shut down on Jul 23, 2026, and computer use has been generally available since GPT-5.4 (Mar 5, 2026).
From Operator to ChatGPT agent
OpenAI launched Operator only for Pro users in the US, planning to expand to Plus, Team and Enterprise and to bring its abilities into ChatGPT later (OpenAI). That happened on Jul 17, 2025, when ChatGPT agent arrived for Pro, Plus and Team users through the tools menu (OpenAI). At that launch Pro users got 400 agent messages a month and other paid users 40.
| Operator (Jan 2025) | Agent mode (Nov 6, 2026) | |
|---|---|---|
| Where | Separate site, operator.chatgpt.com | Inside ChatGPT on web, iOS, Android, macOS and Windows |
| Plans | Pro, US only | Paid plans only |
| Limits | Not stated in the launch post | Plus 40 messages a month, Pro 400, Business and Enterprise 40 |
| How to start | Open the site | Tools menu, or type /agent |
Sources: OpenAI, OpenAI Help Center.
Agent mode in turn sits beside ChatGPT Work, released Jul 9, 2026 as "an agent for longer, more involved tasks". How to use agent mode step by step is on How to Use ChatGPT Agent Mode.
Frequently asked questions
Does OpenAI Operator still exist? No. OpenAI folded Operator into ChatGPT as agent mode on Jul 17, 2025, and the Operator website is no longer accessible.
What replaced Operator? ChatGPT agent mode replaced Operator. Choose it from the tools menu in ChatGPT or type /agent; it is available on paid plans.
When did Operator come out? OpenAI released Operator on Jan 23, 2025, as a research preview for Pro users in the US.
What model did Operator use? Operator ran on the Computer-Using Agent (CUA), which combined GPT-4o's vision with reasoning trained by reinforcement learning to use buttons, menus and text fields.
How many agent messages do I get in ChatGPT? As of Nov 6, 2026, OpenAI's help article lists 40 agent messages a month on Plus, 400 on Pro and 40 on Business and Enterprise; only requests you start count.
Sources
- Introducing Operator, OpenAI, Jan 23, 2025, updated Jul 17, 2025
- Introducing ChatGPT agent, OpenAI, Jul 17, 2025
- ChatGPT agent, OpenAI Help Center, read Nov 6, 2026
- API changelog, OpenAI, read Nov 6, 2026
- Deprecations, OpenAI, read Nov 6, 2026
- ChatGPT release notes, OpenAI Help Center, read Nov 6, 2026
- Get started with ChatGPT Work, OpenAI, Jul 9, 2026 (video)
Moments
Get started with ChatGPT Work
OpenAI
See also
- OpenAIOpenAI is the AI company behind ChatGPT, Codex and the GPT models. Since Oct 28, 2025 it has been OpenAI Group PBC, controlled by the nonprofit OpenAI Foundation, with Microsoft holding about 27%; CEO Sam Altman leads it, and it was last valued at $852 billion.
-
ChatGPTChatGPT is OpenAI's AI assistant, and as of Nov 6, 2026 it is one app with three modes: Chat for questions, Work for longer tasks and Codex for software. It is free to use, with paid plans from $8 to $500 a month. -
Agent ModeAgent mode (ChatGPT agent) lets ChatGPT do tasks on the web in its own virtual browser, such as filling forms or booking, asking before important steps. It launched on Jul 17, 2025, replacing Operator; it is for paid plans, with 40 tasks a month on Plus and 400 on Pro, and longer agentic work now also runs in ChatGPT Work. -
OpenAI APIThe OpenAI API lets programs call OpenAI's models (GPT-6 Astra, GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Luna, image, voice and transcription models) with an API key, billed per token. New projects use the Responses API through the official SDKs, with documentation at developers.openai.com.
-
ChatGPT WorkChatGPT Work is the agent mode of ChatGPT for longer tasks: it researches, works across connected apps and files, and delivers finished documents, spreadsheets, presentations, reports and Sites. It launched on Jul 9, 2026 and shares its usage allowance with Codex. -
How to Use ChatGPT Agent ModeTo use ChatGPT's agent, choose Work for a longer task (or select agent mode from the tools menu or type /agent in a chat), describe the goal, then answer its questions, approve important actions and take over the browser for sign-ins. Agent tasks need a paid ChatGPT plan, as of Nov 6, 2026.