Is ChatGPT Agent the same thing as a Dot?

No. They are different approaches from the same company. ChatGPT Agent refers to the earlier, on-request style where you ask the assistant to carry out a task and it works on it in a session. A Dot is OpenAI's always-on model: you name a primary Dot, it has its own cloud computer and browser, learns from feedback, and can keep working on a project while handling others. See the vendor's current documentation for how ChatGPT Agent is described today.

Can a Dot replace ChatGPT Agent entirely?

It depends on your needs. For ongoing, multi-step work that benefits from continuity and a persistent identity, a Dot is built for that. For a one-off task you want done now, the on-request approach may be simpler. Many people will use both, and OpenAI continues to evolve the two — check its docs for the current feature set.

Do conversations with a Dot count against my ChatGPT limits?

OpenAI says conversations with your Dot do not count toward your ChatGPT usage limits. However, tasks a Dot starts or manages in Codex or ChatGPT Work do count against your existing usage limits.

Which should I use first?

Start from the work. If it is a recurring responsibility you want handled continuously and reviewed, try a Dot. If it is a discrete task you can describe in one sitting, agent mode may be enough. Either way, review consequential output: Dots can make mistakes.

OpenAI Dots vs ChatGPT Agent

It is tempting to treat OpenAI Dots and ChatGPT Agent as the same product at different sizes. They are not. They reflect two different mental models for what an agent is, and understanding that difference is the fastest way to decide which fits a given job.

This page explains the distinction, compares the two on the dimensions that matter, and points to OpenAI's own documentation for the moving parts. It is an independent guide, and the Dots details below come from OpenAI's launch materials — nothing here is invented, and where a detail is not published we say so.

Two mental models

The on-request model is the one most people met first. You describe a task, the assistant works on it — often with a browser or a sandboxed environment — and then it stops. You inspect the result, and if you want more you prompt again. It is a powerful way to delegate a discrete job.

The always-on model is what OpenAI is now selling with Dots. You start with a primary Dot and give it a name. That Dot has its own cloud computer and browser, connects to 4,000+ apps, learns from your feedback over time, and works 24/7. Crucially, it takes on a project and runs with it while handling several others. OpenAI describes Dots reaching you through ChatGPT on desktop, web and mobile, plus Slack and Teams.

The practical upshot: agent mode answers "do this now." A Dot answers "keep owning this." Both are useful; they are not interchangeable.

How the two compare

Dots facts are from OpenAI's launch materials. ChatGPT Agent is described here at a general level because its capabilities change; follow OpenAI's own current documentation before relying on specifics.

Dimension	OpenAI Dots	ChatGPT Agent (on-request)
Runs where	Its own isolated cloud computer with a browser, 24/7	Generally a task environment activated while you work on a request; check OpenAI's docs.
Model	GPT-6 Astra	OpenAI does not fix this in the launch materials; model and behavior change over time.
Identity / credentials	Named primary Dot; specialist Dots have their own identity and systems access	Session-oriented task execution; the identity model differs — confirm in the docs.
App connectivity	4,000+ apps via OpenAI's plugin ecosystem	Depends on your plan and current roll-out; verify against OpenAI's docs.
Approvals / governance	Read-only background research, auto-review, Custom Rules, activity view; can be paused	Typically includes approval or confirmation steps; specifics evolve.
Best fit	Ongoing work that spans channels and needs continuity	A discrete task you can describe and check in one session.
Primary source	OpenAI, Introducing dots	OpenAI's documentation

Interaction model: prompt-and-wait vs ongoing responsibility

The biggest difference is who remembers what. With the on-request model, context tends to live in a conversation and in what you repeat. With a Dot, OpenAI says the agent "learns from feedback over time" and carries context across channels — so the history of the work belongs to the agent, not just to a thread. If your work is a moving target (a launch that keeps changing, an analysis that needs rerunning), that continuity is the whole point.

See OpenAI's safety write-up for how it frames this, including the activity view that shows background work.

Where the boundaries blur — and what to check

OpenAI is one vendor, and its products influence each other. ChatGPT can already start agentic work through Codex and ChatGPT Work — and OpenAI notes that tasks a Dot starts in those surfaces do count toward your ChatGPT usage limits, while conversation with the Dot itself does not. That small footnote is a useful reminder that these are parts of one ecosystem, not rival companies.

Because the labels and capabilities move, treat any specific feature claim as time-sensitive. When in doubt, check the vendor's current documentation rather than a comparison written months earlier.

Which one should you reach for?

It depends on your needs. Reach for the on-request mode when you have a bounded task you can evaluate immediately. Reach for a Dot when the work is ongoing, benefits from persistence, and should reach you wherever you already are. If you are unsure whether a workflow qualifies, our Agent Readiness Scorecard can help you judge it, and Dots pricing covers what an eligible plan includes. Whichever you use, Dots can make mistakes — review consequential work.

Source for Dots facts: OpenAI, Introducing dots. Independent guide; not affiliated with OpenAI.