A practical field note on OpenAI’s Dots: what an always-on agent is, how to create and configure your first one, who can use it today, and where a human should stay in charge.
If you came here from my reel and commented DOTS, this is the guide I promised: what OpenAI’s Dots are, how to create your first one, and the settings worth choosing before you hand it real work.
Read OpenAI’s announcement and the official setup article. Rollout details change, so treat those pages as the final word.
What a dot is
OpenAI introduced Dots on 29 September 2026 as always-on agents inside ChatGPT, powered by GPT‑6 Astra. A normal chat is reactive: you ask, it answers, and the work stops when you close the tab. A dot is designed to keep going.
According to OpenAI, each dot has its own cloud computer and browser, can connect to more than 4,000 apps through plugins, and can work around the clock. You can message it or call it by voice in ChatGPT, and message it from Slack or Microsoft Teams. Texting a dot is a beta for Pro users in the United States.
Reactive assistant versus proactive agent
The useful shift is not a smarter answer. It is ownership of a recurring job.
- Reactive: “Summarise this email thread.”
- Proactive: “Every morning, review my calendar and inbox, flag conflicts, and draft replies for anything waiting on me.”
The second instruction describes an outcome, a schedule and a boundary. That is the kind of work a dot is built for. The examples on this page are illustrations of how you could brief one, not claims about anyone’s results.
Who can use Dots today
At launch, OpenAI said Dots are rolling out to:
- Pro users outside the European Economic Area, Switzerland and the United Kingdom.
- Business Premium workspaces in all regions.
- Enterprise, Edu and Healthcare workspaces as a beta, once an admin enables it.
Your first dot is included at no extra cost. The rollout can take several days, so if you do not see it yet, check again later rather than looking for a workaround.
How to set up your first dot
- Open ChatGPT in the mobile app (iOS or Android), the desktop app (including Windows) or on the web from a desktop browser. Creating a dot is not supported on mobile web.
- Start the dot onboarding and follow the prompts.
- Name your dot. Its handle takes the form @yourname-agentname.
- Choose a character or pet for it. This is its visual identity in ChatGPT.
- Connect the apps it needs from your profile: Customize, then Plugins. Connect Slack and other messaging channels from desktop.
- Set its boundaries under Customize, then Custom rules (next section).
- Give it its first recurring task by asking in plain language, for example “Review my calendar each morning and tell me what needs attention.”
Your dot’s profile shows its work as In progress, Scheduled and Completed, so you can see what it is doing and what it has finished.
Set boundaries before you give it work
Custom rules decide how much a dot may do on its own. For each kind of action you can choose:
- Take action without asking.
- Take action if pre-approved.
- Ask before taking action.
- Hand off to you.
A sensible start is “Ask before taking action” for anything that sends messages, spends money or changes shared documents, and “Take action without asking” only for read-only work such as research and summaries. Loosen the rules after you have reviewed its output for a while.
Local computer access is optional and off by default. If something looks wrong, you can pause your dot from the ••• menu.
How OpenAI describes the safety model
OpenAI says proactive research uses read-only tools, actions pass through an automatic review, and sensitive tasks such as changing a password always stay with you. It also says a dot can make mistakes, so review anything consequential before it goes out.
That matches the practical rule I would give any team adopting agents: automate the preparation, keep a person on the decision.
A starting brief you can adapt
If you want a dot to act like a lightweight chief of staff, a clear first brief might look like this:
Every weekday at 8 a.m., review my calendar, inbox and Slack mentions. Give me a short list of what needs my attention today, draft replies I can approve, and flag any meeting conflicts. Ask before sending anything.
It names the schedule, the sources, the output and the boundary. Start there, review a week of results, then decide what it can do without asking.
Sources & further reading
This page is an independent explainer and does not imply OpenAI endorsement. Availability and features are as described by OpenAI at the time of writing.
Frequently asked questions
What are Dots in ChatGPT?
Dots are always-on agents in ChatGPT, introduced by OpenAI on 29 September 2026 and powered by GPT‑6 Astra. Each dot has its own cloud computer and browser, can connect to apps through plugins, and can work on scheduled tasks around the clock.
Who can create a dot?
At launch, Pro users outside the EEA, Switzerland and the UK, Business Premium workspaces in all regions, and Enterprise, Edu and Healthcare workspaces in beta once an admin enables it. The first dot is included at no extra cost, and rollout can take several days.
Can I create a dot on my phone?
Yes, in the ChatGPT app for iOS or Android. You can also use the desktop app or ChatGPT on a desktop browser. Creating a dot is not supported on mobile web.
Can a dot act without my permission?
Only within the custom rules you set. For each kind of action you can let it act without asking, act if pre-approved, ask first, or hand off to you. Sensitive tasks such as password changes always stay with you.