What is OpenAI dots? The always-on ChatGPT agent, explained

The short definition

OpenAI describes dots as capable, always-on agents that take on ongoing responsibility and keep making progress between conversations (OpenAI). You give your dot a goal and define what it can do on its own. It works through the problem, figures out what needs to happen next, brings results back for your review, and turns to you for decisions that need your judgement (OpenAI Help Center).

The product is called “dots”; each agent is “a dot”. You’ll also see it written as ChatGPT dots, GPT dots or OpenAI dot. It’s all the same thing.

How is a dot different from a regular ChatGPT chat?

You can still ask questions and explore ideas the way you do in ChatGPT; the difference is that the dot keeps following up on work that’s in progress (OpenAI).

What is GPT-6 Astra?

GPT-6 Astra is the model that powers dots. OpenAI calls it its most capable and most aligned model, and says it’s trained to understand your goals, stay within the scope of a request, and ask targeted questions when the answer could affect what happens next. It’s also trained to refuse harmful requests, including biological and cybersecurity misuse (OpenAI).

Its own cloud computer (and, optionally, yours)

Every dot has its own cloud computer where it can browse the web, analyse information, create files and run tools. Each workspace is sandboxed and isolated from other users (OpenAI). You can open your dot’s computer from its profile to check its work or interact with it (OpenAI Help Center).

Access to your own computer is optional and off by default. You connect a computer from the ChatGPT desktop app on that machine and can revoke access at any time. With local access, your dot can create Work or Codex tasks, use local skills, and use your local browser when its cloud browser is blocked (OpenAI Help Center).

Where you talk to it

• ChatGPT on web, desktop and mobile, by message or voice call. Your dot can also message you with progress, questions or decisions (OpenAI).
• Slack and Microsoft Teams. Context carries across channels, so you can start in ChatGPT and add details in Slack (OpenAI).
• Text messages: a limited beta for Pro users in the US, run by a third-party provider. Pro users can also join a waitlist for iMessage and RCS (OpenAI Help Center).

Your dot can’t call you at launch, and it can’t have its own standalone email address.

Proactive research and memory

Even when you haven’t asked for anything, a dot can look through your connected apps for ways to help, for example noticing that your travel plans changed. OpenAI calls this proactive research. It uses read-only tools, so it can’t send messages, change content in your apps, or control a browser or computer. Any follow-up action goes through the normal rules and checks (OpenAI).

Your dot receives memories from ChatGPT and creates its own, including from connected apps (OpenAI Help Center). You currently can’t view or edit individual dot memories; deleting (resetting) the dot is the way to clear its context (OpenAI Help Center). More on this in our safety and privacy guide.

What people use dots for

OpenAI’s launch post gives five example roles (OpenAI):

• Developers: spot repeated requests in customer feedback, scope and build small fixes, test them, and hand over a pull request with a video of the change.
• Launch leads: when product scope changes, work out what that means for the story, and redraft launch materials, assets and docs.
• Scientists: rerun analyses when new data arrives, investigate surprises, update figures, and flag interpretations that need review.
• Sales leads: check customer requirements against product docs, build a proof of concept for a key integration, and keep the proposal and test plan current.
• Content creators: pull clip-worthy moments from a new transcript, prepare show notes, and draft social posts for approval.

Inside OpenAI, dots pick up bugs reported in Slack and turn new designs into working apps. One early tester’s dot noticed an unsent invoice, prepared it and sent it once they approved (OpenAI).

What dots can’t do (yet)

OpenAI is also previewing dedicated dots for organisations, with their own identity and a defined role, starting with targeted enterprise pilots and a Microsoft Agent 365 integration (OpenAI). See availability for the details.