OpenAI launches Dots, proactive agents with their own cloud computer

OpenAI said Tuesday at DevDay it is introducing Dots, proactive assistants powered by GPT-6 Astra that run on their own cloud computer, learn from feedback, and can keep working toward goals around the clock through ChatGPT, Slack, Teams, and connected apps.

Agents that keep working without you babysitting every click are the product pitch of DevDay. Dots put a named agent you can open and inspect, with a computer of its own, next to ChatGPT instead of another chat box.

On Tuesday, 29 September 2026, OpenAI introduced Dots at DevDay. The product post is titled “Introducing dots.” The same day’s DevDay recap covers the launch and links to that post. OpenAI says Dots are always-on agents, and it calls them frontier intelligence powered by GPT-6 Astra. An agent, here, is software that keeps working toward a goal, not only a chat reply. Always-on means it can keep going when you are not in the chat. Each Dot has its own cloud computer, a machine OpenAI runs for it, separate from the person’s laptop. OpenAI says a Dot learns from feedback over time and can work toward a person’s goals 24 hours a day. Those lines are on OpenAI’s post.

OpenAI says Dots connect through its plugin ecosystem to over 4,000 apps. A plugin is a connection that lets the agent use another product. You can reach a Dot in ChatGPT, in Slack, or in Microsoft Teams. You can hop on a voice call when you want to talk something out. OpenAI says texting is coming soon. In ChatGPT you can message or call on a desktop, on the web, and on a phone. You create the first Dot in the ChatGPT desktop app or in a desktop browser. After that setup, OpenAI says you can message it in the mobile app. The more you work together, OpenAI says, the more the Dot learns your preferences, how you think, and what good looks like to you. Those lines are OpenAI’s.

You can open a Dot’s computer at any time and inspect its work. OpenAI says a Dot can connect to other devices besides its own. You can also give it permission to use your laptop, so it can work directly alongside you. Your computer, and what is on it, stays separate unless you choose to connect it. For signing into supported websites, OpenAI says a Dot can use saved passwords without showing those passwords to the model. OpenAI’s example from outside the company: an early tester’s Dot noticed he had forgotten to invoice a publication, prepared the invoice, and sent it after he approved. That example is OpenAI’s.

When you are not actively working with it, OpenAI says the Dot looks for ways to help in the background. OpenAI calls that proactive research. It uses apps you have already connected, and only with read-only tools. That background pass cannot send messages, change what is in an app, or control your browser or your computer. You choose which apps it can reach, through the app controls you already have in ChatGPT. Built-in rules decide when it acts on its own and when it asks. Custom Rules, settings you write, can allow a specific action, require approval, or block it. If monitoring flags a safety concern, OpenAI says it can pause or stop the work. OpenAI also says Dots can still make mistakes, so work that matters should be reviewed. Those lines are OpenAI’s.

Who gets it, as OpenAI states it. The company says it is beginning to roll Dots out across Pro, Business Premium, and Enterprise plans in eligible markets, and that it plans to expand to more users. A later section is more specific. Dots are rolling out in ChatGPT to Pro and Business Premium users in those eligible markets. Enterprise users, including Edu and Healthcare, the education and healthcare workspace plans, can try a beta when a workspace admin turns it on. A beta, here, is a test version that is not on for everyone. The DevDay recap says that enterprise beta is off by default. OpenAI says the first Dot is included in a Pro or Business Premium plan at no extra cost. The plan also includes an allowance for deeper work, with extended limits for the first month after launch. Conversations with the Dot do not count toward ordinary ChatGPT usage limits. If you ask the Dot to start or manage tasks in Codex, OpenAI’s coding product, or in ChatGPT Work, those tasks still count as usual. The post does not name the countries inside “eligible markets.” It does not print a dollar price, or the size of that allowance. Those lines are OpenAI’s.

What OpenAI says comes next. You start with one primary Dot, give it a name, and make it your own. Over time, OpenAI says it envisions teams of Dots working together on your behalf. The same post shares a preview of specialist Dots. A company would set each one up with its own identity, its own credentials, and access to the systems it needs. OpenAI describes that identity as covering access management, hardware the company issues, and deep ties into a company’s systems of record, so the Dot can take on a defined job. A system of record is the software a company treats as the official copy of its data. OpenAI says this is starting as focused enterprise pilots. Its engineers would work with an organization on what the Dot is responsible for, which tools it can use, and how people review and approve the work. OpenAI also says it is working with Microsoft so specialist Dots can sit under Agent 365, Microsoft’s controls for governing and securing agents. The aim is that a business can manage Dots with Microsoft tools it already uses. Those lines are OpenAI’s. They describe a preview and pilots. They are not a claim that every company can turn on a specialist Dot today.

In plain terms, OpenAI is shipping a named agent that keeps working on a computer of its own, reachable in ChatGPT, Slack, and Teams. You can open that computer and watch. Pro and Business Premium users in markets OpenAI calls eligible can start, and the first Dot is included at no extra cost. An enterprise workspace can try a beta if an admin turns it on. Teams of Dots, and specialist Dots inside companies, are what OpenAI says comes next.

The picture is OpenAI’s product screen from the Introducing dots post. It shows a named proactive agent working across connected apps. It is the company’s interface.