OpenAI Dots Explained: What ChatGPT Dots Are, Pricing, Availability and How to Set Up Your Dot

What are OpenAI dots?

OpenAI dots are always-on AI agents inside ChatGPT that take ongoing responsibility for your work. Each dot runs on GPT-6 Astra, has its own cloud computer and browser, works across the apps you connect, and keeps moving toward your goal after you close the chat.

OpenAI announced dots at DevDay on September 29. The company calls them “remarkably capable, always-on agents built to handle everything.” OpenAI writes the name in lowercase, so I do the same here, and I use “dot” for a single agent. Many people also search for “ChatGPT dots”; it is the same product.

I have worked in digital marketing and software since 2012. When a new AI product lands, my first job is to separate launch hype from what the official documents actually say. Therefore, every fact below comes from OpenAI's announcement, its Help Center articles and the ChatGPT release notes. Anything that comes from the press carries a clear “reported in the press” label.

How are OpenAI dots different from a normal ChatGPT chat?

In a classic ChatGPT conversation, you ask, the model answers, and the work stops there. A dot, by contrast, owns a goal. According to the Help Center, a dot “can take on ongoing responsibility and keep making progress between conversations.” In other words, it decides the next step while you do something else.

I group the difference into three points:

• Continuity: A chat is a single exchange. A dot follows work for hours or days and checks in when it needs a decision.
• Workspace: A dot has its own cloud computer, its own browser and the apps you connect.
• Learning: The more you work together, the more it learns your preferences, how you think and what good looks like to you.

A dot can also research in the background on its own. OpenAI calls this “proactive research.” For example, the company describes a dot noticing a change to your travel plans and suggesting how to help. So a dot behaves less like a chat window and more like a digital colleague who works for you.

What is GPT-6 Astra and why does it matter for dots?

GPT-6 Astra is the model that powers OpenAI dots. According to the release notes, OpenAI introduced Astra on September 3 to a limited set of organizations. OpenAI describes improvements in coding, research, computer use and complex multi-step work. Astra can also create documents, spreadsheets and presentations that follow your templates, and it adapts when you change direction.

In its safety post, OpenAI calls Astra “our most capable and aligned model.” Its key skill is understanding your goal, staying within the scope of your request and asking focused questions when the answer could change the plan. For an agent that works on its own for long stretches, that skill is critical. A single misunderstood instruction could lead it to change the wrong file.

In addition, Astra includes extra safety monitoring. If the system detects that the agent may have misread your instructions, it can pause or stop the work as a precaution and let you decide what happens next. In short, the value of dots depends on this brake system as much as on raw intelligence.

Why does each dot have its own cloud computer?

Every dot works on a separate cloud computer that OpenAI maintains. According to the safety post, that environment runs Linux and the Chrome browser. There, the dot can browse, analyse information, create files and run tools. Your own computer and its contents stay separate unless you choose to connect them.

This setup brings two advantages. First, your dot keeps working while your laptop sleeps. Second, sandboxing contains mistakes or harmful code inside that workspace. OpenAI also isolates each user's cloud environment from everyone else's.

You can open your dot's computer at any time to inspect the work. On desktop, you open it from the dot's profile and interact with it directly. On mobile, the computer opens in your control.

You can also connect your own computer, although that access starts turned off. To connect, you use the ChatGPT desktop app on that machine and confirm Allow access. To stop it, you confirm Revoke access. With local access on, your dot can create Work or Codex tasks, use local skills and use your local browser when its cloud browser hits a block.

What do 4,000+ app connections mean in practice?

OpenAI says its dots can connect to more than 4,000 apps through its plugin ecosystem. They can also use the plugins you already connected to your account. In practice, a dot can read your email, calendar, files or team tools and produce work from them.

However, one detail matters a lot. Connected apps feed your dot information in the background too, not only during a task. The Help Center states that your dot can review connected information proactively and form memories from it, even when you haven't asked a specific question.

You can disconnect an app at any time. Still, disconnecting does not delete what your dot already learned from it. The only way to delete that information is to reset the dot. That is why I recommend thinking through each connection during setup. On client projects, I always start with the narrowest access and add apps only when a task truly needs them.

If you use Codex, your dot can also start tasks in Codex cloud environments. You need to create the environment in Codex first.

Where can you reach your dot: ChatGPT, Slack, Teams or a voice call?

You can message your dot or start a voice call in ChatGPT on desktop, web and mobile. You can also reach it in Slack and Microsoft Teams. Your dot carries context across every channel, so a project you start in ChatGPT continues with the same knowledge when you discuss it with the team in Slack.

A few channel rules are worth knowing:

• You set up messaging channels such as Slack from desktop.
• You can connect your personal email account, but at launch you cannot give your dot its own email address.
• Texting is a limited beta for Pro users in the US. It runs through a third-party provider, and message and data rates may apply.
• Your dot cannot call you at launch; you start voice calls yourself.
• To stop outgoing texts, you reply STOP.

As a result, most users outside the US will rely on the ChatGPT apps, Slack and Teams for now.

Which plans include OpenAI dots, and how much do they cost?

With OpenAI dots, your first dot comes with the Pro and Business Premium plans at no extra cost. Enterprise users, including Edu and Healthcare, can try the beta once their workspace admin turns it on. Put simply, you don't buy a separate dots subscription; the first dot ships with an eligible plan.

Usage rules are clear, too. Conversations with your dot don't count toward your ChatGPT usage limits. However, if you ask your dot to start tasks in Codex or ChatGPT Work, those tasks count as usual. The release notes add that, for the next month, dots usage won't count toward eligible Pro, Business and Enterprise plan allowances. After that, OpenAI will share usage terms for each plan.

I also saw headlines describing dots as a $500-a-month product. That reading doesn't match the official text: $500 is the price of the Pro 500 plan, while the first dot comes with every Pro tier.

Are OpenAI dots available in the UK, the EU and other countries?

The Help Center is specific. Dots are rolling out to Pro users “in markets excluding the European Economic Area, Switzerland, and the UK.” So a Pro subscriber in the US, Canada, Australia or Turkey falls within the eligible scope, provided ChatGPT is supported there. Still, the rollout is gradual, and OpenAI says access may take several days to reach your account.

For Europe and the UK, the picture is different. A Pro subscriber in Germany, France or London cannot use dots at launch. However, Business Premium includes dots “across all supported ChatGPT regions.” That means a company in the EU or the UK can access dots through Business Premium seats. On Enterprise, the decision sits with the workspace admin.

There is also an age rule: the release notes say dots are for eligible users aged 18 and older. OpenAI does not explain the reason for the European Pro restriction in its official texts. For that reason, I avoid confident claims such as “it's because of GDPR.” The only confirmed point is that the restriction applies at launch.

How do you set up your first dot, step by step?

You can only create a dot on desktop. You cannot create one in the mobile app, and mobile web doesn't support dots. According to the Help Center, the steps look like this:

1. Open the ChatGPT desktop app (macOS or Windows) or ChatGPT in a desktop browser.
2. Follow the onboarding prompts to create your dot.
3. Give your dot a name. Its default handle is @yourname-dot; once you name it, the handle becomes @yourname-agentname.
4. Choose one of the available characters or a pet. Your dot may also generate a pet for you.
5. Connect your apps and review each app's permissions before you connect it.
6. Set up messaging channels such as Slack from desktop.
7. Let the dot introduce itself. After that, you can message it from the ChatGPT mobile app.

To change the name or avatar later, select the dot's name or avatar to open its profile, then select the pencil icon. I quote the menu labels exactly as the Help Center shows them, because the rollout is still in progress and labels may shift slightly.

How do you give your dot tasks and manage scheduled work?

Giving a task works like messaging a colleague. You describe what you want and share the details it needs. For example, you might ask it to review your calendar, research a topic or remind you about something later. To attach a file or photo, you select + in the conversation.

Your dot's profile shows work in three views: In progress, Scheduled and Completed. For a recurring check, an instruction such as “review my calendar each morning and tell me what's coming up” is enough. Scheduled runs and proactive updates then appear in the conversation.

To manage scheduled tasks, you open Scheduled in the profile. There you review active, paused and completed tasks. Then you open a task to change its repeat schedule, its time or its completion notifications.

Meanwhile, Activity View in the desktop app lists ongoing and delegated tasks, including background work. From there, you can add context, correct a misunderstanding, change direction or ask your dot to stop. That way, the agent never turns into a black box running on its own.

How do connected apps, memory and reset work?

On mobile, you review connected apps by opening your dot's profile and going to Customize → Plugins. This screen uses shared ChatGPT settings. As a result, your connections across ChatGPT, ChatGPT Work and Codex live in one place.

For memory, your dot draws on two sources: the memories ChatGPT already holds about you and the memories it creates itself, including from connected apps. On mobile, Customize → Memory shows what ChatGPT remembers. You can't delete your dot's own memories one by one, though; for that, you need to reset it.

Here is how pause and reset work:

• Pause: Open the ••• menu in the profile and select Pause. To resume, select Paused • Tap to resume.
• Reset: Open the ••• menu, select Reset, review what will go, then select Reset again to confirm or Cancel to keep your dot.
• Effect of a reset: It deletes the dot's conversations, saved memories and scheduled tasks. You then create a new dot on desktop.

How do custom rules and approvals work?

Dots start with built-in rules for when to act on their own and when to ask for approval. On top of that, you can add Custom Rules. On mobile, you open Customize → Custom rules to review the defaults or add a rule. You describe the action the rule covers, then choose one of four behaviours:

1. Take action without asking
2. Take action if pre-approved, where “pre-approved” means you explicitly requested the action in your prompt
3. Ask before taking action
4. Hand off to you

Underneath these rules sits a layer called Auto-review. When a dot plans an action such as sending an email or changing a file, a separate safety system checks the step against your instructions, Custom Rules and safety requirements. For an email, it checks the recipient and the message to catch a wrong address or information you didn't mean to share.

The key point: Custom Rules cannot remove built-in safeguards. Your dot can help you write rules, but it needs your approval to change them. Also, if Auto-review blocks a step, it tells the dot why, and the dot can then ask you, try a permitted alternative, hand the step back or stop.

What safety limits do dots have?

The OpenAI dots safety post makes clear that some actions never happen without you. Knowing these limits helps you calibrate how much you trust your dot:

• Confirmation every time: permanently deleting data, installing or running software from an unrecognised source, and granting new security-sensitive access.
• Handed back to you: changing a password or moving money between financial accounts. The dot helps with the surrounding task, but you do the sensitive step.
• Purchases: A dot can buy with a card you already saved on a merchant's site, but every purchase needs your approval.
• Sensitive data sharing: Health data always needs a named recipient. Less sensitive data, like an email address, needs a class of recipient by default.
• Secure sign-in: On supported sites, a dedicated encrypted credential service passes your password to the browser without exposing it to the model.

In addition, proactive research tasks use read-only tools, and OpenAI enforces that limit in code. Those tasks cannot message other people, change content in connected apps or control a browser or desktop. Against prompt injection, OpenAI combines model safeguards, tool restrictions, checks before actions and monitoring. Even so, OpenAI openly says dots can make mistakes, so you should always review consequential work.

Does OpenAI train on your dot's data, and what about GDPR?

OpenAI says it doesn't use content from ChatGPT Business, Enterprise or Edu workspaces to train its models by default. On personal plans, the choice is yours. If “Improve the model for everyone” is on, your dot's conversations, the work it carries out and the automations you set up may help train models after OpenAI works to remove personal identifiers.

OpenAI doesn't train directly on proactive research threads or your dot's notes to itself. However, if one of those notes enters an eligible conversation, it may then count as part of that conversation for training, depending on your settings. OpenAI also encrypts your content at rest and in transit.

For companies in the EU and the UK, the real question is data protection law. When you connect a CRM, a customer list or a shared inbox, you process personal data through a service provider. Therefore, I suggest these steps:

• Use a Business or Enterprise workspace, which OpenAI excludes from training by default, instead of personal accounts.
• Update your privacy notice and your records of processing.
• Talk to your legal adviser before connecting sources that hold special category data, such as health information.

This is not legal advice. In practice, though, the biggest risk I see is staff connecting company data to a personal account.

How do admins manage dots in a ChatGPT workspace?

In Enterprise workspaces, OpenAI dots arrive switched off. Admins turn access on with the permissions described in the Help Center article on managing dots in workspaces. The main settings are:

Two details deserve attention. First, a dot's cloud computer doesn't inherit a member's local VPN, browser sign-ins or device policies. Second, turning off Custom Rules doesn't make every action require approval; the default action rules still apply. Also, local computer access isn't available in workspaces whose Codex or ChatGPT Work policies target a specific operating system.

What are specialist dots?

Your personal dot works on your behalf. Specialist dots, on the other hand, take on dedicated responsibilities inside an organisation. The company gives each one its own identity, credentials and access to the systems it needs. It can also receive IT-provisioned hardware.

For now, specialist dots are a preview, starting with focused enterprise pilots. OpenAI's engineering teams work directly with each organisation to define the dot's responsibilities, its tools and how people review and approve its work. OpenAI says its internal testing covered procurement, invoice processing, email marketing, customer support and commercial contracting.

OpenAI is also working with Microsoft to bring specialist dots into the governance and security controls of Microsoft Agent 365. The goal is to let businesses manage dots through Microsoft tools they already use. There is no official date yet, so don't treat this integration as a ready feature in your planning.

How can marketing and sales teams use OpenAI dots?

The official examples give a useful frame. Inside OpenAI, a bug appears in Slack and dots immediately start investigating; a new design arrives and dots turn it into a working app. Outside OpenAI, an early tester's dot noticed he'd forgotten to invoice a publication, prepared the invoice and sent it after his approval.

These are the marketing and sales scenarios I find worth testing:

• Weekly reporting: Pull numbers from connected files and sheets and draft a Monday summary. I cover the analysis side in my guide on using AI for data analysis.
• Newsletters and content: As in OpenAI's own example, follow a developing story and prepare a newsletter draft. For the email side, see AI in email marketing.
• Sales follow-up: Draft follow-up emails after meetings, with a Custom Rule set to “Ask before taking action” for sending.
• Competitor and market research: Watch specific sites and flag important changes.
• Team coordination: Track decisions in a Slack channel and remind people about deadlines.

My broader guide on AI agents for marketing helps you decide which jobs to hand to a dot first.

How do dots differ from ChatGPT agent and Operator?

OpenAI has shipped agents before. Operator was its first agent that acted in a browser for you; OpenAI later folded that capability into ChatGPT's agent mode. In agent mode, you start a task, the agent finishes it and the job ends there.

OpenAI dots move that model toward continuity. As reported in the press, the core difference is this: with Computer Use in ChatGPT Work, you start each task individually and the agent operates your own screen. A dot, by default, works on its own computer, hands work to Codex when needed and touches your laptop only with permission.

From the official documents, the differences are also clear. A dot has a name and an identity, learns about you over time, handles scheduled and recurring work, runs proactive research in the background and keeps context across Slack and Teams. Moreover, it can move several projects forward at once without separate threads.

If you want to build your own agents as a developer, the API is the right place to look; my guide on how to use the OpenAI API walks through it.

OpenAI dots vs Meta Muse: what does the press say?

The Meta Muse details in this section come from press reports, not from OpenAI. According to The Decoder, Meta launched its Muse agent about three weeks before dots. The same report says Muse also runs on its own secured computer with a browser, keeps working in the background and asks for approval before sensitive actions.

In short, both products chase the same idea: always-on agents with their own computer that bring a human in for key decisions. If you want a broader accuracy comparison, see my test of ChatGPT, Gemini, Claude, Copilot and Perplexity.

What are the known limits of dots at launch?

Here are the launch limits that appear in official sources:

• Each user starts with one primary dot; more dots will come later.
• You cannot create a dot on mobile, and mobile web isn't supported.
• Your dot cannot call you and has no email address of its own.
• The texting beta is limited to Pro users in the US.
• Information from a disconnected app stays until you reset the dot.
• Pro access excludes the EEA, Switzerland and the UK at launch.
• The rollout is gradual; access may take several days.

On top of that, OpenAI dots can make mistakes, including when they follow your rules. OpenAI advises you to check important details before you rely on a result. I say the same: in the first weeks, give your dot only work you can undo, and read its output regularly.

For the future, OpenAI promises two things: adding more dots and scaling each dot's speed or monthly workload. It hasn't shared prices for these options yet.

What should you do before bringing OpenAI dots into your business?

Before you add any agent to a workflow, write a short readiness list. With my team, we follow this order on client projects:

1. Write down the first three jobs for your dot; each should be measurable and reversible.
2. Connect apps with the narrowest access you can.
3. Start Custom Rules for sending, buying and sharing at “Ask before taking action.”
4. Document the flow of personal data for GDPR or your local privacy law.
5. Use the extended first-month limits as a trial period and log the results.

I cover the wider framework in how to use AI in business. If you wonder how agents will interact with your website, my guide to WebMCP and the agentic web is a good next step.

If you'd like help setting up agents like these, writing the rule set and connecting them to your systems, my team and I can plan it with you through our AI automation services. To build your team's baseline skills first, start with what AI literacy means.

Finally, I recommend reading the sources yourself: Getting started with your dot for setup, the dots safety post for permissions and the ChatGPT release notes for changes. The product is new, so I will keep this guide updated.