OpenAI's dots work 24/7 on their own cloud computer

One early tester's dot noticed he had forgotten to invoice a publication, prepared the invoice and sent it once he said yes. Dots are the always-on agents OpenAI announced on X. They run on GPT-6 Astra and are rolling out today in ChatGPT to Pro and Business Premium users in eligible markets, with an Enterprise beta that workspace admins switch on.

At a glance

• Each dot gets its own cloud computer and browser, reaches over 4,000 apps through plugins, and can be messaged or called in ChatGPT, Slack or Teams, with texting coming soon.
• Custom Rules decide which actions a dot takes alone, which need your approval and which are blocked, while auto-review checks account actions and a password change always stays with you.
• The first dot comes with the plan at no extra cost, but OpenAI gives no figure for the deeper-work allowance or the future price of extra dots and higher speed.

A dot works from its own cloud computer, and your laptop stays out unless you connect it

Each dot works from its own cloud computer, its own browser and the apps you connect, and OpenAI says it can do nearly anything with that kit. You can open the dot's computer at any time to inspect its work. Your laptop stays separate unless you give permission, in which case the dot can connect to it and work directly alongside you.

A dot can run with one project while handling several others, so you keep adding tasks without juggling separate threads. You reach it by message or voice call in ChatGPT on desktop, web and mobile, and in Slack and Teams. It carries context across every channel, messages you with progress, questions or decisions, and learns your preferences the more you work together.

Inside OpenAI, dots investigate Slack bugs and turn designs into working apps

OpenAI describes its own use: a bug appears in Slack and dots start investigating, a new design arrives and dots turn it into a working app while the team focuses on customer feedback, and a planning cycle stays in sync toward its deadline. Today you start with one primary dot, which you name and make your own. Over time OpenAI envisions teams of dots working together.

A preview of specialist dots targets organizations. Each gets its own identity for access management, credentials, IT-provisioned hardware and deep integrations with a company's systems of record. OpenAI builds on internal tests across procurement, invoice processing, email marketing, customer support and commercial contracting, starts with focused enterprise pilots where its engineers define each dot's duties with the customer, and is working with Microsoft to bring specialist dots under Agent 365 governance and security controls.

Background work runs on read-only tools, and changing a password stays with you

When you are not working with it, a dot looks for ways to help on its own, which OpenAI calls proactive research. It uses your connected apps only through read-only tools, which cannot send messages, change app content or control your browser or computer. For supported websites, a dot can sign in with saved passwords without exposing them to the model.

Acting is gated separately. Dots start with built-in rules on when to act and when to ask, and Custom Rules let you allow, require approval for or block specific actions. Before anything that could affect your accounts or share information, auto-review checks it against your instructions, those rules and fixed safety requirements, deciding what proceeds, what needs your approval and what you must do yourself.

Think of a new assistant with a badge that opens the filing room but not the payments desk, plus a supervisor who reads each outgoing letter. A monitoring system watches for harmful behavior and malicious instructions and can pause or stop the dot. Activity View shows progress, including background work, so you can redirect it.

Chatting with a dot does not count toward ChatGPT usage limits

Your first dot is included in Pro or Business Premium at no extra cost, with an allowance for deeper work and extended limits for the first month after launch. Conversations with the dot sit outside ChatGPT usage limits, but tasks it starts or manages in Codex or ChatGPT Work count as usual. Later you will be able to add more dots and scale each one's speed or monthly workload.

Enterprise users, including Edu and Healthcare, get a beta once a workspace admin enables it, and OpenAI plans to reach more users soon. Setup starts in the ChatGPT desktop app or a desktop browser, where you connect apps and let the dot introduce itself. After that you can message it from the ChatGPT mobile app.

On training, content from Business, Enterprise and Edu workspaces is not used by default, and personal-plan users can decide whether dots' conversations and work improve models. OpenAI does not train directly on proactive research or a dot's notes to itself, though that material may inform an eligible conversation or task depending on your settings.

Oddly for a product pitched on taking work off your plate, a practical number is missing: the plan includes an allowance for deeper work, yet OpenAI states neither its size nor what the extended first-month limits fall back to. The announcement also admits dots can still make mistakes and asks you to review consequential work, and it points to a system card rather than giving capability figures for GPT-6 Astra.

When the first-month limits end

The extended limits cover only the first month after launch, and OpenAI does not say what the allowance looks like afterwards. Texting, extra dots and scaling a dot's speed or workload are described as coming soon or in the future, with no dates. Specialist dots remain in focused enterprise pilots, and no timeline has been given for the Microsoft Agent 365 integration.