OpenAI dots are always-on AI agents that run in OpenAI’s cloud, work across your apps around the clock, and come included with a ChatGPT Pro or Business Premium plan. You give a dot a name, connect the apps you trust it with, and it takes on ongoing work: researching, drafting, tracking tasks, and reporting back. Think of it less as a chat window and more as a junior teammate who never sleeps.
That is the short answer. The rest of this guide covers what a dot actually is, what it costs, what it can do, what stays off limits, and whether it is worth turning on.
What exactly is a dot?
A dot is an agent, not a chatbot. The difference is that an agent takes actions on its own rather than waiting for you to type every step.
Each dot gets its own cloud computer and its own browser. It can open files, browse the web, run tools, and work on several projects at once. You can open its computer at any time to inspect what it is doing. OpenAI calls them “always-on agents that take on ongoing responsibility.”
They run on GPT-6 Astra, OpenAI’s current flagship model. One naming note, because headlines this week were confusing: GPT-6 Astra powers dots. GPT-6.1 Sol is a cheaper model OpenAI announced the same day. GPT-6.1 Astra is a different model OpenAI shelved the day before over safety concerns. They are three separate things.
What does it cost?
The first dot is free with your plan. There is no separate line item for one dot on Pro or Business Premium.
- Pro: $200 per month in the US.
- Business Premium: about $100 per seat per month.
- Enterprise, Edu, and Healthcare: a beta, once a workspace admin enables it.
- Extra capacity: OpenAI says you will later be able to add more dots or scale each dot’s speed and monthly workload for an additional cost. That pricing is not published yet.
Two billing details that matter. Conversations with your dot do not count against your ChatGPT usage limits. But tasks the dot starts in Codex or ChatGPT Work do count, as usual. And there is a new $500-per-month Pro tier that adds “Ultrafast” output; separately, from October 30, 2026, the existing $200 Pro plan takes a usage cut in Work and Codex.
What can a dot actually do?
OpenAI’s own examples are a decent guide:
- Watch a Slack channel and start investigating a reported bug.
- Turn a new design into a working app while the team does something else.
- Keep a project moving toward a deadline and nudge the people involved.
- Notice a missed invoice, prepare it, and send it after you approve.
- Run background research and bring you findings you did not ask for.
It connects to more than 4,000 apps through the existing ChatGPT plugin ecosystem. You can reach it in ChatGPT on desktop, web, and mobile, and in Slack and Teams. Texting is “coming soon,” and voice calls work.
What stays off limits?
This is the part OpenAI is selling hardest, and it is genuinely more guardrailed than most agents.
- Isolation: each dot runs on its own cloud computer, separate from yours, unless you choose to connect your laptop.
- Read-only background work: “proactive research” uses tools that cannot send messages, change app content, or control a browser. This is enforced in code.
- Auto-review: a separate system checks planned actions against your instructions and rules before they run, and its controls live outside the dot’s environment, so the dot cannot switch them off.
- Kill switch: monitoring can pause or stop a dot mid-task and show you a warning.
- Hard handoffs: changing a password, moving money between accounts, permanently deleting data, or installing software from an unrecognized source always comes back to you. Some of those require confirmation every single time.
- Passwords: a secure sign-in flow pauses the model while you log in, so credentials are not handed to the model.
You manage what a dot can reach through your existing ChatGPT app controls and Custom Rules. There is also an Activity View where you can watch background work and redirect it.
Where is it available?
Dots are rolling out now in eligible markets. There is one big gap: Pro users in the UK, the European Economic Area, and Switzerland cannot get dots yet. Business Premium does include dots in all supported regions, which makes it the only route in for those countries for now. OpenAI has said it plans to expand access over time.
How is this different from just using ChatGPT?
ChatGPT answers when you ask. A dot acts when you are not looking.
Plain ChatGPT is a conversation: you bring a question, you get a response. A dot carries context across sessions, runs in the background, connects to your apps, and can start and finish multi-step work. It also lives in your team’s tools, not just a browser tab. If you have used an agent like OpenClaw or Meta’s Muse, dots is OpenAI’s version of that same idea, hosted in its own cloud.
What about enterprise?
Beyond your personal dot, OpenAI is piloting specialist dots: dedicated agents with their own identities and credentials that take on defined jobs inside a company, such as procurement or invoice processing. They are in pilot with engineering support, and a Microsoft Agent 365 integration is in progress so businesses can manage them through Microsoft’s existing controls.
Should you use one?
If you already pay for Pro or Business Premium, turning one on costs nothing extra, and the guardrails are the most detailed of any consumer agent so far. Let it handle low-risk research and drafting first, connect only the apps you actually want it to touch, and watch the Activity View for a week before you trust it with anything that matters.
Two honest caveats. The safety story is OpenAI’s own, described by OpenAI, on OpenAI’s cloud, and the company had a rough month of agent incidents right before this launch. And the more work you push to a cloud agent, the less of it runs on hardware you control. Start small, and decide deliberately what you hand over.
The Short Version
- Dots are always-on agents in OpenAI’s cloud, powered by GPT-6 Astra, with their own computer and browser.
- Your first dot is free with Pro ($200/mo) or Business Premium ($100/seat). More dots and more capacity will cost extra later.
- They work in ChatGPT, Slack, and Teams, and connect to 4,000+ apps. Texting is coming soon.
- Strong guardrails: isolation, read-only background work, auto-review, a kill switch, and hard handoffs for sensitive steps.
- Not available to Pro users in the UK, the EEA, or Switzerland yet.
For the head-to-head breakdown, see OpenAI’s dots vs OpenClaw. And if you go the self-hosted route, our guide to monitoring an OpenClaw agent covers the failure modes to watch for.
FAQ
Do I need the $500 plan to use dots? No. Dots are included with Pro and Business Premium. The $500 tier adds faster output and more capacity, not access to dots themselves.
Is my data used to train OpenAI’s models? Not by default for Business, Enterprise, and Edu workspaces. On personal plans, you control it through your data settings. OpenAI says it does not train directly on proactive research or a dot’s notes to itself.
Can a dot spend my money or send messages on its own? Only within rules you set. Messages and file shares require authorization that matches the data and the recipient, and purchases with saved cards require approval. Broad actions can be allowed through Custom Rules, but some sensitive steps always come back to you.
Can I run dots on my own computer? No. Each dot runs on an OpenAI cloud computer. You can optionally connect your laptop, but that reduces the isolation and some protections.
When will dots reach the UK and EU? OpenAI has not given a date. Business Premium works in supported regions today, and OpenAI says it plans to expand access over time.
New to always-on agents in general?
Dots is not the only one. Meta’s Muse arrived first and has its own strengths and problems. If you want the plain-English version of what these agents are before you pick one, start there.
About The Author
