TellWorldAI World Archive

Home›Agents

Agent

Operator

Operator is an AI agent OpenAI released on 23 January 2025 that works in a web browser by looking at screenshots and using a mouse and keyboard. Its abilities were folded into ChatGPT agent in July 2025.

Era Generative AI for everyone

Release date
Developer
OpenAI

What it is

Operator is an AI agent that performs tasks in a web browser. OpenAI released it on 23 January 2025 as a research preview, and at launch it was available only to ChatGPT Pro users in the United States[1][2]. An agent in this sense is a system that carries out a sequence of actions for a user, rather than only producing text.

Origins

Operator ran on the Computer-Using Agent (CUA), a model that combined GPT-4o's vision with reasoning trained through reinforcement learning[3][6]. Reinforcement learning is a training method in which a model is rewarded for behaviour that achieves a goal. The CUA model worked from screenshots of the browser and acted with a mouse and keyboard, in the way a person would use a computer[3][6].

OpenAI also offered the CUA model to developers as computer-use-preview through the Responses API from 11 March 2025[3][5].

Main features

Operator could fill in forms, place online orders and schedule appointments. It ran in a virtual machine with limited internet connectivity, which kept its activity inside a contained environment[4]. It struggled with complex user interfaces, meaning websites with many steps or unusual layouts were harder for it to handle[4].

Operator was trained to ask for confirmation before it completed tasks such as booking a hotel or sending an email[3]. Agents of this kind differ from older scripted browser automation in that a model chooses each click and keystroke from what it sees on the screen. This check meant that the user, not the agent, made the final decision on actions with real-world consequences.

Its place in development

The CUA model was shut down in the API on 23 July 2026[3][5]. In July 2025 Operator's capabilities were folded into ChatGPT agent, which controls a virtual computer in the same way[4].

For readers, Operator is a clear early example of an AI system that uses a computer the way a person does. Its history also shows one path such products took: a limited preview for paying users, a model offered to developers, and then a merger into a broader agent.

20253

  1. ReleaseOpenAI launches ChatGPT agent

    OpenAI's ChatGPT agent, launched on 17 July 2025, carries out multi-step tasks on its own virtual computer — browsing websites, running code and producing editable files — combining Operator's web interaction with deep research's ability to gather and summarise information.

  2. ReleaseOpenAI launches the Responses API and Agents SDK

    On 11 March 2025 OpenAI released the Responses API and an open-source Agents SDK, merging the simplicity of Chat Completions with the tool use of the Assistants API and bundling web search, file search, computer use and tracing.

  3. ReleaseOpenAI launches Operator, a web-browsing agent

    On 23 January 2025 OpenAI released Operator, a research preview agent that used its own browser to fill forms, place orders and schedule appointments for ChatGPT Pro users in the US. It combined GPT-4o vision with reinforcement learning and introduced confirmation and takeover safeguards.

Developed by1

OpenAI

Powered by1

GPT-4o

Notes and sources

  1. OpenAIhttps://en.wikipedia.org/wiki/OpenAI
  2. OpenAI GPT Model Release Timelinehttps://hidekazu-konishi.com/entry/openai_gpt_model_release_timeline.html
  3. OpenAI unveils AI agent Operatorhttps://forklog.com/en/openai-unveils-ai-agent-operator/amp
  4. ChatGPThttps://en.wikipedia.org/wiki/ChatGPT
  5. Deprecationshttps://developers.openai.com/api/docs/deprecations
  6. Introducing Operatorhttps://openai.com/index/introducing-operator/

Cite this page: Operator. TellWorld, https://www.tellworld.net/agents/operator. Updated 2026-11-08