OpenAI used its DevDay 2026 keynote on Tuesday to announce a bunch of new features, including Dots. It’s basically a new kind of AI agent built to run continuously in the background rather than wait for you to open a chat window.
These always-on AI agents (or Dots) run on GPT-6 Astra, which, if you don’t already know, is OpenAI’s most powerful AI model.
So what can a dot actually do?
A Dot runs on its own dedicated cloud computer. It can connect to over 4,000 apps through OpenAI’s plugins, which let it respond and take action rather than just chat (which is essentially the difference between AI agents and chatbots).
OpenAI also gives us a few examples in its official blog post. A Dot can investigate a bug the moment it’s mentioned in Slack, turn a finished design into a working app, or notice an unsent invoice and prepare it for approval.
When you’re not delegating a Dot, it does “proactive research” through read-only connections, which is a safety measure that prevents it from sending messages or taking action without your permission.
How do I create Dots, and who can use them?
To create one, start in the ChatGPT desktop app or desktop browser, name your Dot, and connect your apps. Through your ChatGPT app controls, you can set Custom Rules to specify which actions your Dot can perform independently, which require your manual approval, and which are completely blocked.
After that, reach it through ChatGPT, Slack, or Teams, and manage it from mobile too.
Dots roll out today to ChatGPT Pro and Business Premium subscribers in eligible markets. OpenAI includes one Dot for free, with more available to add later. Enterprise, Education, and Healthcare users get it once their admin enables it.
Conversations with your Dot don’t count against usage limits, so that’s a good thing. However, all the tasks it runs through Codex or ChatGPT Work do. OpenAI isn’t hiding the target here: it launched days after Meta’s Muse dominated app charts, and it even borrows its bubbly, cartoonish persona.
OpenAI isn’t blazing a new trail here, though. Meta’s Muse, xAI’s Grok Bots, and Anthropic’s Claude Cowork are already pushing AI beyond the chat window and into more autonomous, always-on work.
