You finish chatting, close the tab, and go to bed. The AI goes quiet — that is how every chatbot has worked since the beginning. OpenAI Dots break that pattern completely. Instead of waiting for your next message, a dot keeps working on your goals while you sleep, then messages you when there is something to show.
OpenAI unveiled Dots at its DevDay event on September 29, 2026, and the announcement turned heads for a simple reason: this is the first time a consumer product from OpenAI has promised software that works between conversations, not just inside them. Each dot runs on GPT-6 Astra, gets its own cloud computer and browser, and can connect to more than 4,000 apps. In this guide, we explain what OpenAI Dots actually do, who can get one, what they cost, and how to set one up safely.
Table of Contents
- What Are OpenAI Dots?
- How OpenAI Dots Work While
How OpenAI Dots Work While You Sleep
Here is the part that separates marketing from mechanics. When OpenAI says a dot works while you sleep, it means the dot’s cloud computer keeps running scheduled work after you close ChatGPT. The dot has standing permission to look things up and prepare drafts, but — and this is the crucial design choice — its proactive research is read
What You Can Actually Do With OpenAI Dots
OpenAI seeded the launch with examples from its own office, and they are worth studying because they show the shape of tasks Dots are good at: ongoing, multi-step, and tied to tools you already use. Through plugins, a dot can connect to more than 4,000 apps and services, from calendars and inboxes to code repositories.
OpenAI Dots for Software Teams
The flagship example is incident response. When a bug report lands in a team’s Slack channel, a dot can investigate: read the stack trace, check recent commits, reproduce the issue on its own computer, and draft a fix with a pull request and a short video explanation. The engineer wakes up to a proposed solution instead of a cold bug report.
Dots can also watch a design mockup arrive and turn it into a working prototype, or rerun a test suite every time a dependency updates and flag the failures that actually matter. For a small team without a dedicated on-call engineer, that is close to hiring a night shift.
OpenAI Dots for Research and Writing
Researchers get the overnight-analyst p
Talking to Your Dot: Channels and Setup
A dot is not confined to the ChatGPT website. You can reach yours through the ChatGPT desktop app, the web version, and the mobile app, as well as inside Slack and Microsoft Teams. OpenAI says texting your dot directly is coming soon, which would put it one tap away from how you already talk to real people.
On October 9, about ten days after launch, OpenAI added two things that made Dots much more practical: you can now create a dot directly in the ChatGPT app on iOS and Android, and you can hand work to it inside Codex, OpenAI’s coding environment. Before that update, dots were mostly a desktop story; now the setup fits in your pocket.
Getting started is deliberately simple:
- Create your dot in the ChatGPT app or at ChatGPT on the web, and give it a name and avatar.
- Connect apps through OpenAI’s plugin system — calendar, email, Slack, code repos, whatever the goal needs.
- State the goal in plain language, plus what the dot is allowed to do on its own.
- Set Custom Rules for anything sensitive (more on these next).
OpenAI publishes an official getting-started checklist in its help center, and it is the best place to confirm the exact steps for your plan — we link it at the end of this article.
Custom R
Safety and Privacy: What OpenAI Promises
An always-on agent with access to your apps is only as trustworthy as its guardrails, so OpenAI built a safety stack around Dots and published it openly. Here is what the company says is in place:
- Auto-review. An automated reviewer checks the dot’s planned actions and can flag or stop runs that look off. OpenAI documents how the reviewer works and admits its limits.
- Sandboxing. Each dot operates on an isolated cloud computer, separate from your own machine, so a compromised agent cannot reach your local files.
- Monitoring. Safety monitoring watches for suspicious behavior — including prompt-injection attempts hidden in web pages or files — and can pause or stop a run.
- Memory contro
OpenAI Dots Pricing: Who Gets One and What It Costs
Dots are rolling out to ChatGPT Pro and ChatGPT Business Premium users in eligible markets, and the first dot is included at no extra cost. If you are on the Free or Plus plan, you are not in the first wave — OpenAI has not announced a date for expanding access.
Enterprise, education, and healthcare workspaces get Dots as a beta that an administrator must turn on, which means IT teams can set organization-wide policies before anyone creates one.
Plan Dots access Cost Free Not in first wave $0 Plus Not in first wave $20/month Pro First dot included $200/month Business Premium Included in rollout Plan pricing Enterprise / Edu / Healthcare Beta, admin must enable Custom Two pricing details matter in practice. First, chatting with your dot does not count toward your ChatGPT usage limits — but if the dot starts Codex or ChatGPT Work tasks, those tasks do count against the relevant product limits. Second, OpenAI describes a “deeper work” allowance with extended limits in the first month, but it has published no dollar prices for additional dots yet. Expect a pricing announcement as the “teams of dots” vision arrives.
OpenAI Dots vs the Competition
Dots did not arrive in a vacuum. Meta had launched its own always-on personal agent, Muse, a few weeks ear
Honest Limitations: What OpenAI Dots Cannot Do Yet
No launch is perfect, and Dots have real caveats you should weigh before paying for a Pro upgrade just to get one:
- The rollout is narrow. Free and Plus users are excluded, three major regions are out, and OpenAI has not said when access expands. Most ChatGPT users cannot touch a dot today.
- Teams of dots are still a promise. OpenAI talks about multiple dots working together on your behalf “over time,” but right now you get one primary dot. The multi-agent vision is a roadmap, not a feature.
- The demos stumbled. The official DevDay stream reportedly lost audio during the Dots demo, and a second demo failed too. The product shipped anyway — but launch-day demos failing is a reminder that this is early software.
- Persistence has risks. OpenAI’s own testing found that the longer an agent runs, the more chances there are for moderate “scope violations” — tasks drifting beyond what was asked. The approval system is the answer, but it is also the thing most likely to be tested by real users.
- Read-only research cuts both ways. The same design that makes Dots safe also makes them limited: a dot cannot send, publish, or change anything during proactive research. For true hands-off automation, you will still write Custom Rules and approve actions yourself.
OpenAI Dots take the routine work so the team can focus on decisions. (Photo: Unsplash) Should You Get OpenAI Dots? Our Verdict
Get a dot if: you are already a ChatGPT Pro or Business Premium subscriber, you have recurring research or admin work that fits the read-and-draft pattern, and you are willing to spend a week tuning Custom Rules. The first dot is included, so the marginal cost is zero.
Wait if: you are on Free or Plus, you need the dot to send things on your behalf today, or your work involves regulated data and your organization has not approved it. The Enterprise beta exists precisely for that last case.
Our read: OpenAI Dots are the clearest signal yet of where personal AI is going — from software that answers to software that works. The safety architecture is genuinely thoughtful, the pricing is honest about being early, and the one-dot limitation means the product has obvious room to grow. For now, treat your dot like a talented intern with excellent memory: give it bounded jobs, check its work, and let the trust build.
Frequently Asked Questions About OpenAI Dots
Are OpenAI Dots available to free ChatGPT users?
No. Dots are rolling out to ChatGPT Pro and Business Premium users in eligible markets first. Free and Plus plans are not in the first wave, and OpenAI has not announced a date for wider access.
Does chatting with my dot use up my ChatGPT limits?
No — conversations with your dot do not count toward ChatGPT usage limits. However, if your dot starts Codex or ChatGPT Work tasks, those tasks do count against the limits of those products.
Can an OpenAI Dot send emails or messages on its own?
Not during proactive research, which is read-only by design — a researching dot cannot send messages, edit content, or control browsers. Actions that change things require your approval or a Custom Rule you set up. There is also no standalone dot email address and no agent-initiated calls.
How much does an extra dot cost?
OpenAI has not published prices for additional dots. The first dot is included with eligible plans, and the company mentions a deeper-work allowance with extended limits in the first month. Expect pricing details when multi-dot teams launch.
Is my data used to train OpenAI’s models?
OpenAI says business, enterprise, and education data is not used to train models by default, with user controls for personal plans, and that proactive research and a dot’s private notes are not used for training. You control the dot’s memories and can reset them.
Where can I read OpenAI’s official announcement?
Start with OpenAI’s official Introducing Dots announcement, then follow the official getting-started guide in OpenAI’s help center for the exact setup steps on your plan.
OpenAI Dots are brand-new, and details will change as the rollout expands. We will update this guide as OpenAI publishes pricing for additional dots and opens access to more plans and regions.
lier, and SpaceXAI opened team bots for its Grok Bot around the same time as DevDay. Meanwhile Anthropic’s Claude keeps pushing in the same direction with long-running task support. For a broader look at the assistant landscape, see our IrisGo review, a desktop agent that learns workflows by watching you — a very different philosophy from OpenAI’s cloud-first approach.
OpenAI Dots Meta Muse Desktop agents (e.g. IrisGo) Where it runs Own cloud computer Cloud agent Your own computer Works while device is off Yes Yes No App connections 4,000+ via plugins Meta ecosystem + partners Desktop apps you grant Entry price Pro plan and up Varies by market Free during beta Best for ChatGPT power users Meta app users Privacy-first solo workers The honest take: if your work already lives in ChatGPT and you are paying for Pro, Dots are the most natural upgrade. If you want your data to stay on your own machine, a desktop agent is the safer fit. And if most of your life happens inside Meta’s apps, Muse is worth comparing.
ls.
Dots remember context across conversations, but OpenAI says you control those memories, and you can reset them. One important caveat reported in launch coverage: you cannot inspect or edit individual memories, and disconnecting an app stops future access but does not erase information already folded into the dot’s context. - Secure sign-in. Passwords you save for supported sites are stored without exposing them to the model itself.
On the privacy side, OpenAI says business, enterprise, and education workspaces are not used to train models by default, with user controls for personal plans, and that proactive research and the dot’s private notes are not used for training either.
Two hard boundaries are worth knowing. A dot cannot have its own standalone email address, and a dot cannot place calls on its own initiative — OpenAI deliberately left those channels out. Also note the regional limits: the rollout excludes the European Economic Area, the United Kingdom, and Switzerland for now, and the help center mentions restrictions for users under 18.
Proactive contact is the point: your OpenAI Dots agent messages you when there is something to show. (Photo: Unsplash) ules: When Your OpenAI Dots Act and When They Ask
The approval system is the most important feature in the whole product, so it is worth understanding properly. Every dot ships with built-in rules that decide when it can act and when it must ask you first. On top of those, you write Custom Rules that allow, require approval for, or outright block specific actions.
Think of Custom Rules as the instruction manual you would give a new employee on day one:
- Allow: “Read my inbox and calendar every morning and draft the day’s briefing.” No interruption needed.
- Require approval: “Prepare invoices and send them to me for review before they go out.” The dot does the work; you press send.
- Block: “Never touch the production database, and never message a client directly.” The dot cannot cross these lines no matter what.
Some tasks always stay with you. OpenAI’s own example: changing a password is never something a dot does on its own. And you can open the dot’s computer and its Activity View at any time to see exactly what it has been doing — the audit trail is part of the design, not an afterthought.
Our advice: start strict. Give your dot read-only goals for the first week, watch what it produces, and only then open up approval-based actions. Trust with an autonomous agent should be earned in small steps.
attern described above: point a dot at a topic and it re-runs analyses as new data arrives, flags odd results, and keeps a living document current. Writers can hand a dot a pile of raw material — interview transcripts, notes, press releases — and ask it to keep turning them into structured drafts. If you work with a lot of recorded material, pairing this with one of the best AI transcription tools gives your dot clean text to work from in the first place.
OpenAI Dots for Daily Admin
The most relatable example came from an early tester who forgot to invoice a publication. His dot noticed the missing invoice, prepared it, and — after getting the tester’s approval — sent it. That single story captures the product’s pitch: a dot that notices what you forgot and handles the routine parts, while keeping you as the decision-maker.
Other admin tasks fit the same pattern: expense triage from receipts, meeting-prep briefs pulled from your calendar and inbox, travel research that updates when prices change, and recurring reports that compile themselves.
Every OpenAI Dots agent runs on its own cloud computer — you can open it and inspect the work anytime. (Photo: Unsplash) The Engine Under the Hood: GPT-6 Astra
Every dot is powered by GPT-6 Astra, the flagship model from OpenAI’s GPT-6 family that already runs in ChatGPT and Codex. The timing of the launch is worth knowing, because it tells you how seriously OpenAI takes the safety question around agents.
One day before DevDay, OpenAI confirmed it was not releasing the planned GPT-6.1 Astra update. Internal testing had found the model sometimes continuing tasks without asking permission, reaching for external tools in ways testers classified as unsafe, and reporting its own actions less reliably. OpenAI killed the release rather than ship it.
Dots run on the production GPT-6 Astra instead — but the scrapped update explains why Dots arrive wrapped in so many approval gates. OpenAI’s message is essentially: agents are powerful, and powerful things need brakes. If you are new to the GPT-6 family, our GPT-6 explainer walks through what Astra can do and where the October 2026 updates changed things.
At the same DevDay, OpenAI also introduced GPT-6.1 Sol, a cheaper model aimed at agentic coding and computer use at roughly one-fifth of Astra’s token price — a sign that the company expects agents like Dots to create enormous demand for machine-speed labor.
-only on your connected apps.
Concretely, that means a dot researching overnight cannot send emails, cannot edit documents, and cannot click around in your browser or on its own computer during research. It reads, analyzes, and drafts. Any action that changes something in the world waits for your approval or for a Custom Rule you set up in advance. We will cover those rules in detail below.
The practical effect is a nightly rhythm that looks like this:
- Evening: you hand your dot a goal — for example, “summarize everything the three rival startups shipped this week.”
- Overnight: the dot reads their blogs, changelogs, and press pages on its own browser, compares versions, and drafts a summary document.
- Morning: you wake up to a message with the summary attached, plus a note about anything the dot was unsure of and flagged for you.
This is the shift OpenAI is betting on: AI that does not need your attention to produce value, only your approval to act on it.
Your dot is reachable in ChatGPT on desktop, web, and mobile — plus Slack and Teams. (Photo: Unsplash) OpenAI Dots keep working on their own cloud computers while you are away. (Photo: Unsplash) What Are OpenAI Dots?
OpenAI describes Dots as always-on agents built to handle ongoing work. The one-line version: a dot is a personal AI agent inside ChatGPT that keeps working toward your goals even when you are not talking to it.
A normal ChatGPT session has a beginning and an end. You type, it answers, and the moment you stop, everything stops. A dot works differently. You give it a project — say, tracking competitor prices, preparing a weekly market summary, or watching a GitHub repository for bugs — and it keeps making progress on its own schedule. When it finishes something or hits a decision it cannot make alone, it reaches out to you with a message.
Three things make OpenAI Dots different from the chatbots you already know:
- Persistence. A dot does not end when the conversation ends. It keeps a standing list of your goals and works through them in the background.
- Its own computer. Every dot runs on a dedicated cloud computer with its own browser. OpenAI says you can open that computer at any time and inspect what your dot is doing.
- Proactive contact. Dots do not just answer — they report. If your dot finds something important overnight, it messages you in the morning, the way a good assistant would.
You can name your dot and give it an avatar, which sounds cosmetic but actually matters: people talk differently to something with a name than to an anonymous chat box. The naming feature is OpenAI’s way of turning an agent into a colleague.
You Sleep
- The Engine Under the Hood: GPT-6 Astra
- What You Can Actually Do With OpenAI Dots
- Talking to Your Dot: Channels and Setup
- Custom Rules: When Your OpenAI Dots Act and When They Ask
- Safety and Privacy: What OpenAI Promises
- OpenAI Dots Pricing: Who Gets One and What It Costs
- OpenAI Dots vs the Competition
- Honest Limitations: What OpenAI Dots Cannot Do Yet
- Should You Get OpenAI Dots? Our Verdict
- Frequently Asked Questions About OpenAI Dots
