At a glance
| What it is | An AI agent that works toward a goal around the clock |
|---|---|
| Made by | OpenAI, inside ChatGPT |
| Announced | September 29, 2026, at DevDay |
| Runs on | GPT-6 Astra, on its own computer in the cloud |
| Reaches | More than 4,000 apps through ChatGPT plugins |
| Talk to it in | ChatGPT, Slack or Microsoft Teams |
| Who can get one | ChatGPT Pro and Business Premium; a beta for Enterprise, Edu and Healthcare |
| Price | The first dot comes with the plan; extra dots aren’t priced yet |
What a dot is
Until now, ChatGPT mostly waited for you. You asked, it answered, and when you closed the tab the work stopped. A dot turns that around. You give it a goal that takes days rather than seconds, like keeping a project’s status page current or chasing replies to a batch of invoices, and it keeps at it on a computer of its own in OpenAI’s cloud. It doesn’t run on your Mac, and it doesn’t stop when you close the lid.
It also doesn’t just report back at the end. You can open a dot’s computer while it works and watch what it’s doing, which matters more than it sounds: an agent you can watch is one you can correct before a small misunderstanding becomes a sent email. Under the hood every dot runs on GPT-6 Astra. (OpenAI writes the name in lowercase: dots, not Dots.)
How you set one up
- Give it a goal. Describe the outcome you want, not a list of clicks. The dot works out the steps.
- Connect your apps. Dots reach more than 4,000 services through ChatGPT’s plugins, from email and calendars to Slack and project tools. It can only touch what you connect.
- Set its rules. For each kind of action you choose one of four settings, the ones in the planner above. This is the part worth slowing down for.
You do this in the ChatGPT desktop app or on the web on a computer. At launch, dots aren’t in the ChatGPT phone app. Once a dot is running, it can reply to you in ChatGPT, in Slack or in Microsoft Teams, so it can report wherever you already are.
The four permission levels
Every kind of action a dot might take gets one of four settings. On its own means exactly that. If pre-approved means it can act within limits you set up front. Ask me first means it stops and waits for you each time. Hand to me means it doesn’t try at all: it gets things ready and passes control back.
Some lines OpenAI draws for you. When a dot researches in the background, it is read-only: it can look through your connected apps, but it can’t message anyone from there. Changing a password or transferring money always comes back to you, whatever you’ve set. And approval doesn’t stack. Saying yes to one email doesn’t give a dot permission to send the next; each one is asked again. Before any action runs, an automatic review checks it against your instructions and rules, such as who an email is going to and what it says.
A sensible way to start is the way the planner opens: let it read and draft freely, ask before anything leaves your accounts, and loosen it one kind of action at a time as it earns it.
Who can use dots, and what they cost
Dots are rolling out gradually to ChatGPT Pro and Business Premium subscribers aged 18 and over. On Pro, the European Economic Area, Switzerland and the UK are excluded at launch; Business Premium covers every region ChatGPT supports. Enterprise, Edu and Healthcare workspaces get a beta, switched off until an admin turns it on. The Free, Plus and Business plans aren’t included.
Your first dot comes with the plan. OpenAI says you’ll be able to add more dots and buy extra speed or monthly capacity later, but hasn’t said what that will cost. Talking to a dot doesn’t count against your normal ChatGPT usage, though work it does through Codex or ChatGPT Work counts against those products’ own allowances.
What dots can’t do yet
- Have their own email address. A dot sends from the account you connect, or not at all.
- Run from the ChatGPT phone app. Setup and management are desktop-only for now.
- Reach Pro subscribers in the EU, Switzerland or the UK. There’s no date yet.
- Act during background research. That mode only reads.
OpenAI has said more is coming, including texting a dot and teams of dots that work together, and specialist dots for jobs like procurement and customer support are being piloted with businesses.
Is it safe to let one loose?
OpenAI published an unusually candid number with the launch. In its own testing, the share of tasks flagged for moderate “scope violations”, meaning a dot doing more than it was asked, rose from 8.6% to 19.7% as the number of unrelated tasks in between grew from five to ten. It saw no severe breach or data leak in that test, and it says vulnerabilities remain. Put simply, the longer and busier a dot’s day, the more it drifts. That’s the case for the “ask me first” setting, and for watching its computer now and then.
Dots and Meta’s Muse
Three weeks before dots, Meta launched Muse, a personal agent that also works on its own computer in the cloud. The two are built for different lives. Dots live inside ChatGPT and Slack and are sold with paid work plans; Muse lives in WhatsApp and Meta’s apps, is free to start, and is aimed at errands like bills, travel and shopping. How Meta’s Muse works.
Keeping an eye on your dot
The one habit that makes an agent safer is looking at it. If you use Tabula, open ChatGPT as a tile on the canvas next to the work your dot is helping with, so its progress is in view instead of behind a tab you forget to check. That’s not a dots feature, just a canvas being useful. What a spatial browser is.
Questions
- What is ChatGPT dots?
- Dots are always-on AI agents inside ChatGPT, announced by OpenAI on September 29, 2026. Each dot runs on its own computer in the cloud, connects to your apps, and keeps working toward a goal you give it even when you are not in ChatGPT.
- Is ChatGPT dots free?
- No. At launch dots are part of the ChatGPT Pro and Business Premium plans, which include your first dot at no extra cost. OpenAI says it will sell additional dots and more capacity later but has not published prices.
- Can I use dots in the UK or the EU?
- Not on a Pro plan yet: Pro access excludes the European Economic Area, Switzerland and the UK at launch. Business Premium is available in all of ChatGPT’s supported regions.
- Can a dot spend my money?
- Not on its own. Transferring money is one of the actions where OpenAI makes a dot hand control back to you, along with changing a password. For everything else you choose, per kind of action, whether it acts alone, acts within limits you pre-approve, asks first, or hands off.
- Do dots use up my ChatGPT limits?
- Talking to a dot does not count against your normal ChatGPT usage. Work a dot starts or manages in Codex or ChatGPT Work counts against those products’ own allowances.
- Can I use a dot on my phone?
- Not from the ChatGPT mobile app at launch. You create and manage dots in the ChatGPT desktop app or on the web on a computer; OpenAI says mobile support will come in an update. A dot can also reply to you in Slack or Teams.