OpenAI Dots: What They Are, How They Work and What They Cost

OpenAI Dots are always-on AI agents inside ChatGPT. You give a dot a job, it keeps working in the background on its own cloud computer, and it comes back to you for approval before anything consequential happens. Dots are powered by GPT-6 Astra and can reach more than 4,000 apps through plugins.

OpenAI announced them at DevDay on September 29, 2026. One dot is included at no extra charge for ChatGPT Pro and Business Premium in eligible markets, but the price of additional dots has not been disclosed.

This guide covers access, delegation, safety controls, exclusions, cost, early problems and the comparison with Meta’s Muse. Everything is current as of September 30, 2026, one day after launch, so expect details to shift. For the wider event, see our OpenAI DevDay 2026 announcements guide and the OpenAI DevDay 2026 recap.

What Are OpenAI Dots and How Are They Different From ChatGPT?

OpenAI Dots are persistent agents that live in ChatGPT and keep working toward your goals between conversations. OpenAI’s own Getting started with your dot article calls a dot “an always-on agent in ChatGPT that can take on ongoing responsibility and keep making progress between conversations.”

A normal ChatGPT chat waits for your next message. A dot does not. It has its own dedicated cloud computer and browser, so it can research, draft and check things while your laptop is closed. It also learns your preferences and working habits, and remembers past conversations and decisions.

You pick a name for your dot and a blob or animal-style avatar. The mascot is cute, but the pitch is serious: Sam Altman told CBS News you can delegate work “the way you would to a high agency engineer or a chief of staff.”

OpenAI Dots reach you through the ChatGPT mobile app and web, plus Slack and Microsoft Teams. Texting and phone voice are listed as coming soon. If you already use ChatGPT inside team chat, our guide to the ChatGPT Slack and Teams integration covers the related @ChatGPT feature, which is separate from dots.

How Do You Get Access to OpenAI Dots?

You get access by being on ChatGPT Pro or Business Premium in an eligible market, then creating your dot in the ChatGPT desktop app or on desktop web. Enterprise, Edu and Healthcare workspaces get a beta once an admin switches it on. Access can take a few days to appear.

Setup has one catch. The help center says you cannot create a dot on mobile, although you can talk to an existing one there. OpenAI also says dots work best on desktop.

Step 1: Check your plan and market

Confirm you are on Pro or Business Premium and that your country is on OpenAI’s eligible-markets list.

Step 2: Create the dot on desktop

Open the ChatGPT desktop app or desktop web and find dots in the sidebar under “New chat.” Give it a name. Its default handle is “@yourname-dot.”

Step 3: Connect only what the first task needs

Your dot can connect to Slack and to a personal email account, and OpenAI says you can also give it permission to connect to a laptop. Connect the narrowest set of apps first, then widen later.

Dots sit on top of the same plan tiers that gate GPT-6 Astra. Our guide to how to access GPT-6 Astra in ChatGPT explains the plan tiers that sit underneath.

How Do You Assign Work to an OpenAI Dot?

You assign work by describing the outcome you want and sharing the details the dot needs, in plain language. The help center’s instruction is to “describe what you want your dot to do and share the details it needs.” You can attach files or photos with the plus button, and reach the same dot from Slack or Teams.

Good first tasks are small and checkable. OpenAI’s examples include reviewing your calendar, researching a topic and setting reminders.

Two habits help. State the standard, not only the task, for example “flag anything from a customer, ignore newsletters.” And run a task once before you schedule it to repeat.

Your dot’s profile shows work under three headings: In progress, Scheduled and Completed. This is the Activity View, and you can step in from there. You can also pause a dot from its menu.

Are OpenAI Dots Safe? Custom Rules, Activity View and Auto-Review

OpenAI Dots are built with layered controls, but they are not risk-free. The controls are Custom Rules that decide what needs your approval, an Activity View to watch and redirect background work, an auto-review check on consequential actions, read-only proactive research and an isolated cloud computer for every dot.

Here is what each one does, based on OpenAI’s Dots announcement and the help center:

For organisations, the workspace admin controls matter more. The Manage dots in ChatGPT workspaces article says access is off by default for Enterprise. Owners can separately control local computer access, custom rules, Slack and Teams, cloud browser use, cloud network access, cloud computer use and password manager access. It also says Enterprise model controls and default model settings do not apply to dots.

One day before DevDay, OpenAI withheld GPT-6.1 Astra because it “didn’t quite meet the bar in terms of staying within scope and authorization,” according to its head of safety systems. Dots run on GPT-6 Astra, not that model, but scope discipline is what an always-on agent needs, so test the guardrails rather than assume them.

Where Are OpenAI Dots Available, and Who Is Excluded?

OpenAI Dots are available to Pro and Business Premium users in “eligible markets,” with Enterprise, Edu and Healthcare in beta when an admin enables it. Fortune reports they are unavailable in the European Economic Area, Switzerland and the UK, citing strict data privacy laws. As of September 30, 2026, no dated rollout has been announced for those regions.

Sources differ on detail. OpenAI’s announcement does not list the eligible markets and points to a help article instead. Some coverage says only the Pro tier is excluded in Europe and the UK, with Business Premium reaching more places. Check your account rather than relying on one report.

For L&D teams in the UK and EU, who often face the strictest learner-data rules, that means watching from the sidelines for now.

How Much Do OpenAI Dots Cost?

Your first dot costs nothing extra: it is included in ChatGPT Pro and Business Premium. OpenAI has not disclosed the price of additional dots, faster dots, higher monthly output or specialist dots. So the honest answer to “what does a dot cost” is the price of your plan, plus an unknown for anything beyond one.

The usage rules are also unsettled. The help center says the plan includes “an allowance for deeper work, with extended limits for the first month after launch.” Reports from Fortune and dev.to say conversations with your dot do not count against normal ChatGPT usage limits. OpenAI has not published the numerical allowance for intensive work.

ChatGPT Pro is the $200 tier, and The Next Web reports its Work and Codex allowance falls from 20x to 10x Plus from October 30, 2026. The new Pro 500 tier costs $500 per month. One commentary piece cites $500 per month for Business Premium, but I could not confirm that on OpenAI’s pages, so get a quote before you budget. Our ChatGPT pricing guide tracks the current tiers.

What Are the Limitations of OpenAI Dots So Far?

Early limitations are real. Every’s Dan Shipper reported in a hands-on Vibe Check frequent permission errors, dropped messages and unreliable browser connections, and advised anyone who dislikes rough edges to wait “a week or two.” Beyond bugs, dots cannot phone you and have no email address of their own.

Other gaps come from the documentation and reviewers:

• Memory is all or nothing. Resetting a dot deletes its conversations, saved memories and scheduled tasks. Reviewers say you cannot view or delete individual memories.
• Dots do not replace shared team tools. A dot has no ownership assignment or team hand-offs, so a shared support inbox still belongs in a help-desk platform.
• No independent evaluations exist yet. Everything about reliability comes from OpenAI and a handful of early testers.

OpenAI itself says dots can make mistakes and advises reviewing consequential work first. That is sensible advice. For what the underlying model can do, see OpenAI Astra computer use. For teams building their own agents rather than using dots, see the OpenAI Agents API with computer use guide.

OpenAI Dots vs Meta Muse: How Do They Compare?

Dots and Meta’s Muse are close rivals: both give you a named, mascot-style agent with its own cloud computer and browser that keeps working in the background and asks before sensitive actions. The differences are pricing, audience and channels. Muse launched September 8, 2026 with a free tier, while dots launched September 29 inside paid plans.

Muse details come from Axios on the Muse launch and TechCrunch on Muse for Small Business.

Muse has had its own trouble. Semafor reports that Amazon blocked it from shopping and that it accidentally disclosed a user’s home address. Fortune notes that both products gather a lot of personal data as they personalise. Neither is proven at scale.

For a training team, the practical difference is governance. Dots come with Slack and Teams reach, workspace admin switches and a planned Microsoft Agent 365 link. Muse is aimed first at individuals and small businesses.

What Can You Use an OpenAI Dot For? Five Practical Examples

An OpenAI dot suits work that is ongoing, spread across several apps and easy to check. These five examples come from OpenAI’s launch material and early hands-on reports. They show what is intended, not guaranteed results, and none has been independently tested at scale.

1. Inbox and Slack triage. Shipper used his dot to filter messages and surface what mattered.
2. Calendar watching. A dot can flag clashes and draft a rescheduling note for your approval.
3. Refreshing research. OpenAI’s examples include updating an analysis when new data arrives.
4. Proposal and launch material updates. OpenAI mentions revising launch materials as scope changes, and keeping sales proposals current.
5. Bug fixes from customer feedback. OpenAI describes developers using dots to prepare tested fixes and pull requests.

Dots can also work inside ChatGPT Space, the shared team hub. For more model-level ideas, see our GPT-6 Astra use cases.

Can OpenAI Dots Help Learning and Development Teams?

Possibly, though nobody has shown it yet. OpenAI has published no L&D examples, and I could not confirm a connector for a specific LMS. If your LMS exports reports or has an app in the plugin directory, a dot could take over two chore-heavy jobs below. Treat both as pilots, not proven results.

Could a Dot Chase Course Completions?

A dot could read a scheduled completion report from your LMS, find learners approaching a compliance training deadline, and draft reminder messages to those learners or their managers. With Custom Rules set to ask first, nothing goes out until a training coordinator approves the batch.

That could trim manual follow-up for mandatory courses. It would not replace Kirkpatrick-style evaluation of learning, and learner records should stay out of scope until legal and IT agree the data flow.

Could a Dot Draft SME Interview Prep?

A dot could gather an SME’s calendar slot, existing decks, past project notes and your draft learning objectives, then prepare a briefing and a list of interview questions before the call. The instructional designer reviews it, adjusts it and runs the conversation.

The dot handles preparation, not the SME’s knowledge, and it can misread source material. For more on the model in this field, see our guide to GPT-6 Astra for learning and development.

Conclusion

OpenAI Dots are a serious first step toward agents that carry work between conversations, but they are one day old. The controls look thoughtful, the price of anything beyond your first dot is unknown, and early testers hit bugs. That mix argues for a careful trial rather than a company-wide rollout.

If you are on Pro or Business Premium in an eligible market, create a dot on desktop, connect one low-risk source and give it one checkable task for a week. Keep Custom Rules on ask-first, review the Activity View daily and write down every miss.

If you are in the UK or EU, or on Plus, revisit in a few weeks. Meanwhile, read the DevDay 2026 announcements guide to see which other launches affect your stack.