OpenAI introduced ChatGPT dots on September 29, 2026, and they are agents that keep working on your behalf after you close the app. Each dot runs on its own cloud computer, powered by GPT-6 Astra, and can reach more than 4,000 apps through OpenAI’s plugin ecosystem.
That changes the shape of the product. Until now ChatGPT answered when you asked. A dot is meant to notice things, do the research, and come back with something finished, which is a different kind of trust to hand over.
TL;DR: ChatGPT dots are always-on agents that run on a separate cloud computer and work across ChatGPT, Slack, Teams, and voice calls. They are rolling out to Pro and Business Premium users first, with the first dot included at no extra charge. Sensitive actions go through auto-review and your own rules, and some tasks, like changing a password, always stay with you.
How ChatGPT dots work when you are not looking
A dot is not a chat thread. According to OpenAI’s announcement, it works toward your goals around the clock, runs several projects at once, and learns your preferences and standards from feedback.
The examples OpenAI gives are practical ones: turning customer feedback into tested code fixes, rerunning an analysis when new data lands, or revising marketing material when a project’s scope shifts.
Background research uses read-only access to your connected apps. That detail matters, because it means the agent can look around without being able to change anything while it does. It also raises the question of who gets to use one first.
Who can use ChatGPT dots right now
Access is narrow at launch, and the ChatGPT release notes say dots are rolling out gradually to users aged 18 and older.
| Group | Access | Cost of first dot |
|---|---|---|
| Pro | Rolling out in eligible markets | Included |
| Business Premium | Rolling out in eligible markets | Included |
| Enterprise, Edu, Healthcare | Beta, enabled by workspace admin | Not stated |
| Plus and Free | Not listed | Not stated |
OpenAI says future expansion and stronger capabilities will carry additional pricing. If you are weighing the top tiers, our look at the ChatGPT Pro 500 plan covers what the $500 a month tier adds.
What stays under your control
OpenAI says each dot works from an isolated cloud computer, separate from your own devices. Passwords are handled without being exposed to the model, and an auto-review step checks actions against your instructions and safety requirements.
You set Custom Rules to allow an action, require your approval, or block it outright. An Activity View lets you follow progress, including work done in the background.
Some tasks never leave your hands. OpenAI states that sensitive tasks, such as changing a password, always stay with you. Not a dramatic promise, but a sensible line to draw first.
Where dots show up on your phone
Setup starts on desktop or web, and mobile messaging works through the ChatGPT app after that. SMS texting is listed as coming soon, so for now the phone is a place to check in, not to configure anything.
Dots can also talk to you through Slack, Teams, and voice calls. Conversations with your dot do not count toward ChatGPT usage limits, but tasks it hands to Codex or ChatGPT Work do.
That split is easy to miss, and it decides how fast a busy dot eats into your allowance. It also connects to a recent ChatGPT change on the personal side, which we covered in the ChatGPT Health tab update.
What this means for how you use ChatGPT
If you already run a Pro or Business Premium account, the first dot costs nothing extra, so the real cost is attention. You will be deciding which apps it can read, which actions need your approval, and what you are comfortable letting it finish alone.
For everyone else, this is a preview of where assistants are heading. Google is reshaping its own custom agents in Gemini, and a background agent with its own computer is the next step in the same race.
Start with rules that require approval for anything that sends, spends, or deletes. You can loosen them later, once the Activity View has shown you how the dot actually behaves.
