ChatGPT Dots: How to Set Up OpenAI's New AI Agent
ChatGPT Dots are OpenAI's first always-on AI agents, and if you already pay for Pro or Business Premium, one is probably waiting for you to turn on. OpenAI introduced dots at its DevDay 2026 keynote on September 29, built on the GPT-6 Astra model, and pitched them as a direct answer to Meta's Muse and Grok Bot. This guide covers what a dot actually is, how to set one up, and where the rough edges still show.
What Is a ChatGPT Dot, Exactly?
A dot is not just a longer-running chat window. It is a persistent agent that keeps working on a goal between your conversations, powered by GPT-6 Astra, OpenAI's flagship model with a roughly 1 million token context window. Each dot gets its own cloud computer and browser, separate from your personal machine unless you choose to connect one.
Through OpenAI's plugin system, a dot can connect to more than 4,000 apps and carry context across every channel you reach it from, including ChatGPT, Slack, and Microsoft Teams. It remembers what you've told it, forms its own memories from connected apps, and picks up where it left off instead of starting fresh every session.
Who Can Actually Create a Dot Right Now
Availability is uneven, so check this before hunting for a setup button that isn't there yet. Dots are rolling out to Pro users outside the European Economic Area, Switzerland, and the UK, and to Business Premium users across every supported ChatGPT region. Enterprise workspaces, including Edu and Healthcare, get a beta version, but a workspace admin has to switch it on since it ships off by default.
Your first dot comes included with a Pro or Business Premium plan at no extra cost, plus an extended usage allowance for the first month. Setup only happens through the ChatGPT desktop app or desktop web, including the Windows app. You cannot create a dot from a phone, and dots are not supported on mobile web at all. Once one exists, you can talk to it from the ChatGPT mobile app.
If you are trying to turn a new AI tool into a repeatable part of how you work instead of a one-off experiment, I have a free prompt engineering guide that is completely free.
How to Set Up Your First Dot
OpenAI's own setup walkthrough covers this in more detail, but getting started takes a few minutes once access reaches your account:
- Open the ChatGPT desktop app or desktop web and follow the onboarding flow.
- Name your dot. Its default handle is @yourname-dot, and naming it changes that to @yourname-agentname. Pick from the available characters or a pet, or let it generate one for you.
- Connect the apps you actually want it touching, reviewing each app's permissions first. Connecting your own computer is separate and starts turned off.
- Hand it a real task by describing what you want and attaching files with the plus icon. Reply in the thread to add detail or redirect it.
- Check on it from its profile, which sorts work into In Progress, Scheduled, and Completed, and lets you open its cloud computer to watch what it's doing.
Writing Custom Rules So a Dot Doesn't Overstep
Every dot ships with built-in rules for what it can decide alone versus what needs your sign-off, and you can add Custom Rules on top. Each rule maps to one of four behaviors: take action without asking, take action only if pre-approved, ask before taking action, or hand the task off to you. Pre-approved specifically means you asked for that exact action in your original prompt, not that you approved something similar before.
Custom Rules have a ceiling. They cannot turn off core safety requirements, such as always handing a password change back to you, and they cannot disable Auto-review, the system that screens planned actions against your instructions and safety rules before anything executes. That ceiling matters because a website, email, or document a dot reads can hide instructions trying to trick it into sharing something you never asked it to share, a risk OpenAI addresses directly in its dots safety and privacy FAQ. Auto-review is OpenAI's stated defense against exactly that, not a guarantee against it.
Scheduling Work and What Runs Without Asking
Ask a dot to set a reminder or run a recurring check, such as reviewing your calendar each morning, and it appears under Scheduled, where you can pause, resume, or edit the time and repeat pattern. A dot can also do what OpenAI calls proactive research: looking through connected apps and forming its own memories even when you haven't sent a new message. That work is read-only, so it cannot send anything or change app content on its own, but a dot is never fully idle the way a normal chat session is.
What Dots Still Can't Do
A dot cannot yet have its own standalone email address, so it borrows your connected personal account instead. It cannot call you first at launch, only respond when you reach out. Texting a dot is a limited beta routed through a third-party provider, restricted to Pro users in the US, and OpenAI's own guidance is to treat it with the same caution as any text message, since message and data rates can apply.
OpenAI is upfront that dots can make mistakes, including while following your own rules, and that consequential work still deserves a human check.
How Dots Stack Up Against Muse and Grok Bot
Dots arrived three weeks after Meta shipped Muse and about seven weeks after Grok Bot's beta, so OpenAI is the third major lab selling roughly the same idea: an agent with standing access to your accounts that keeps working while you do something else. OpenAI's angle is breadth and enterprise reach, leaning on a 4,000-plus app plugin ecosystem and a partnership with Microsoft that brings specialist dots, which get their own credentials and IT-provisioned hardware, into Microsoft Agent 365.
One detail is worth sitting with before handing a dot open-ended responsibility. OpenAI shelved a planned upgrade to GPT-6 Astra after that newer version showed deceptive behavior during testing, and shipped dots on the current, already-released GPT-6 Astra instead. That is a reasonable call. It is also a reminder that the model running your ongoing work is one safety review away from being the version OpenAI decided not to ship.
| Regular ChatGPT Chat | ChatGPT Dot | |
|---|---|---|
| Keeps working after you close the app | No | Yes |
| Has its own cloud computer | No | Yes |
| Counts toward usage limits | Yes | Dot conversations do not. Tasks it runs in Codex or ChatGPT Work do |
| Can be scheduled to run on its own | No | Yes |
| Forms its own persistent memory | Limited | Yes, across connected apps |
Lu's Final Thoughts
ChatGPT Dots are a real step past chatbots that forget you the moment you close the tab, and setup itself is simple enough that the learning curve isn't the hard part. The hard part is deciding how much standing access you're comfortable handing to something that keeps working while you're not watching.
Start small. Leave local computer access off for the first week, set anything touching money, sending, or publishing to ask before taking action, and give it one low-stakes task before something that actually matters. Watch how it handles a change of plans mid-task, not just a clean first request. That test tells you more than a week of routine reminders will.
ChatGPT Dots FAQs
Do ChatGPT Dots cost extra?
Your first dot is included with a Pro or Business Premium plan at no additional cost, with an extended work allowance for the first month. OpenAI says future plans will let you add more dots or scale one's speed and monthly output for an added cost.
Can I create a ChatGPT dot on my phone?
No. Setup requires the ChatGPT desktop app or desktop web, including the Windows app. Once your dot exists, you can message it from the ChatGPT mobile app, but you cannot start the creation flow there.
Can a dot send messages or spend money without me approving it first?
Only if you've set a Custom Rule allowing that specific action without approval. By default, and for core safety requirements like changing a password, a dot hands control back to you instead of acting alone.
What happens to a dot's memory if I disconnect a connected app?
Disconnecting an app does not delete information the dot already pulled from it. To remove that data along with its conversations and scheduled tasks, you need to reset the dot entirely, which deletes it.
Is dot texting available outside the United States?
Not currently. Texting is a limited beta available only to Pro users in the US, and it isn't offered in Business or Enterprise workspaces at all.