How to Use OpenAI Dots: A Step-by-Step Guide From Setup to Custom Rules
You have access to OpenAI dots, now what? A practical guide to creating your first dot, connecting apps, scheduling tasks and setting custom rules.
Contents 11
OpenAI's dots, announced at DevDay, turn ChatGPT from a chat window you ask questions into a teammate that keeps working on your behalf. We covered what dots are and the GPT-6 Astra model behind them in an earlier post. This one is practical: how do you set up your first dot, how do you hand it work and, most importantly, how do you decide what it does on its own and what it asks you about?
In short:
- Your first dot can only be created in the ChatGPT desktop app or on desktop web; on mobile you can only talk to it.
- A dot has its own cloud computer and browser; access to your own computer is off by default.
- With Custom rules you pick one of four behaviors per action: act without asking, act if pre-approved, ask first, or hand off to you.
- The first dot is included in Pro and Business Premium at no extra cost, and conversations with it don't count against your ChatGPT usage limits.
Who can use dots?
According to OpenAI's Help Center, dots are rolling out gradually to ChatGPT Pro users in markets excluding the European Economic Area, Switzerland and the UK. Business Premium users get them in all supported regions. In Enterprise, Edu and Healthcare workspaces the beta starts turned off, and an admin has to enable it.
Access can take several days to reach your account. If you don't see the option yet, that isn't a bug; the rollout is happening in waves.
How do you create your first dot?
You need a desktop. Dots are created in the ChatGPT desktop app (there is a Windows version too) or in ChatGPT in your computer's browser. You can't create one from the mobile app, and mobile web isn't supported at all.
- Create the dot. Follow the onboarding prompts; the dot will introduce itself and ask what it can help with.
- Name it. The default handle is
@yourname-dot. Once you name it, it becomes@yourname-agentname. You can change the name and avatar later with the pencil icon on its profile. - Pick a character. Choose one of the available characters or a pet, or let your dot generate one for you.
- Connect your apps. A dot is only as useful as what it can see. Without your calendar, email or Slack, it's a new hire with no context.
- Set up messaging channels on desktop. Slack and texting are connected from desktop too. After setup you can talk to your dot from the ChatGPT mobile app.
How do you give a dot a task?
Giving a dot work is like messaging a colleague: describe what you want and share the details it needs. Use the + button in the conversation to attach a file or photo.
The real difference is that the work doesn't end when the chat does. In classic ChatGPT, closing the window stops everything. A dot keeps the goal in mind, figures out the next step and comes back when it needs your judgment. So a good task description has three parts:
- Goal: not "write a report" but "prepare the Q3 sales summary for Friday's board meeting".
- Definition of done: "One page, three charts, compared with last quarter."
- Boundary: "Don't send anything to the client; leave it as a draft."
A sample first task:
Every morning at 8:30, check my calendar. For each person
I'm meeting that day, summarize our latest conversations
and note one open issue I should remember. Don't send any
email, just send me the summary.
Your dot's profile groups work into In progress, Scheduled and Completed. From the profile you can also open the dot's cloud computer to watch what it's doing, and even take over.
How do scheduled tasks work?
To set a reminder or a recurring check, just ask: "Every Monday morning, list the unpaid invoices." Scheduled runs and proactive updates show up in the conversation.
The Scheduled section of the profile lists active, paused and completed tasks, where you can change the repeat schedule, time and completion notifications.
A dot can also work without a new message from you. OpenAI calls this proactive research: while you're away, the dot looks through the apps you connected for ways to help. The key detail is that this background work only uses read-only tools, so during proactive research a dot can't send messages, change app content, or control your browser or computer.
Custom rules: you draw the lines
The key to using dots safely is Custom rules. A dot starts with built-in rules that decide when it acts independently and when it asks for approval, and you can add your own for specific actions. When writing a rule you describe the action it covers and choose one of four behaviors:
| Behavior | What it means | Example |
|---|---|---|
| Take action without asking | The dot just does it | Adding a note to your calendar |
| Take action if pre-approved | Only if you explicitly asked for that action in your prompt | Sending an email after "send this draft to Alex" |
| Ask before taking action | Waits for your approval every time | Spending money, sharing files |
| Hand off to you | Never does it, leaves it to you | Signing a contract |
"Pre-approved" means you explicitly requested the action in your prompt. On mobile these settings live under Customize → Custom rules.
Our advice for getting started is simple: mark everything that spends money, sends information outside or can't be undone as Ask first or Hand off to you. Loosen the rules as the dot earns your trust. Some sensitive tasks, such as changing a password, are always left to you anyway.
Can a dot use my own computer?
Yes, but only if you want it to. Each dot has its own cloud computer, and your computer stays separate from it. Local access is off by default.
To connect a machine, choose Allow access in the ChatGPT desktop app on that computer. From then on, the dot can reach files and work on that computer from whichever channel you message it. With local access on, it can create Work or Codex tasks, use local skills, and fall back to your local browser when its cloud browser is blocked on a site. Revoke access turns it off at any time.
A note for developers: a dot can also create cloud tasks in Codex cloud environments, as long as you create the environment in Codex first.
Limitations worth knowing
The launch version has a few things to keep in mind:
- No email address of its own. You can connect your personal email account, but you can't give the dot a standalone address.
- It can't call you. You can voice-chat with it, but it can't initiate a call.
- Texting is a limited beta. For now it's only open to some Pro users in the US, and it runs through a third-party provider, so be careful with sensitive information.
- Disconnecting doesn't make it forget. When you disconnect an app, information the dot already got from it isn't deleted. The only way to remove it is to reset the dot.
- Codex and Work tasks count against limits. Chatting with your dot doesn't, but tasks it starts in Codex or ChatGPT Work count as usual.
Pausing and resetting a dot
If you want your dot to stop for a while, use Pause in the ••• menu on its profile. Tap "Paused • Tap to resume" to start it again.
Reset is permanent: it deletes the dot along with its conversations, saved memories and scheduled tasks. Afterwards you create a new dot from desktop again.
Where can businesses use dots?
OpenAI's own examples show dots shine in work that repeats but needs follow-through: a dot that watches customer feedback, fixes small bugs and opens review-ready PRs; one that updates launch materials as the scope changes; one that refreshes a proposal as requirements move. OpenAI also describes an early tester whose dot noticed they had forgotten to invoice a client, prepared the invoice and sent it after getting approval.
What these examples share is telling: the dot prepares the work and asks a human at the critical step. Aim for the same balance when you set up yours.
On the enterprise side, OpenAI offers a limited preview of "specialist dots" with their own identities and defined responsibilities, and is working with Microsoft to integrate them with the governance controls in Microsoft Agent 365.
Frequently asked questions
Is using a dot extra cost?
The first dot is included in Pro and Business Premium at no extra cost. The plan also includes an allowance for deeper work, with higher limits in the first month after launch. OpenAI says you'll later be able to add more dots and scale each one's speed or monthly workload.
Can a dot make mistakes?
Yes. OpenAI says plainly that a dot can make mistakes, even while following your rules. Review anything with real consequences, and keep your rules strict for the first few weeks.
What's the difference between a dot and a normal ChatGPT chat?
A normal chat runs when you type and stops when you close it. A dot takes on responsibility: it keeps making progress between conversations, remembers context, runs scheduled tasks and comes back only when a decision needs you.
Dots try to turn AI from a tool into a team member, and that calls for a new skill: defining tasks well and drawing good boundaries. If you're considering agent integrations or workflow automation built around your business, reach us through our enterprise software development page.
