If you're building with (or around) OpenAI's new always-on agents, this is the setup and permissions walkthrough for dots. It's researched from OpenAI's own documentation, and the sources are linked throughout.
TL;DR: To use OpenAI dots you need ChatGPT Pro ($100, $200 or $500 a month, outside the UK, EEA and Switzerland), a ChatGPT Business Premium seat, or an Enterprise workspace where an admin has switched dots on. Plus and Go don't qualify. As of October 10, 2026, you can create your dot in the ChatGPT mobile app as well as on desktop. Before you give it real work, connect only the apps it needs, set plugin permissions to read-only where you can, add custom rules for sending and deleting, and learn the difference between Pause, Activity and Scheduled, because pausing a dot doesn't cancel its recurring tasks.
Last updated: October 10, 2026.
OpenAI launched dots on September 29 as always-on agents inside ChatGPT. This guide covers how to use OpenAI dots safely once one shows up in your account: setup, what to connect, and which switches to flip before you hand it your inbox.
Everything here comes from OpenAI's own documentation on learn.chatgpt.com, checked on October 10, 2026, plus reporting where noted. We haven't run a dot ourselves, so this is a researched walkthrough of the documented setup, not a review of how well dots perform.
What a dot actually is
A dot is a personal agent that keeps working between your conversations. According to OpenAI's Meet dots page, it runs on GPT-6 Astra, has its own computer and browser in the cloud, and can keep working while your own devices are off.
You give it ongoing responsibilities rather than one-off prompts, such as keeping an offsite plan current. It decides when to follow up, comes back to you for decisions, and can hand jobs to Codex or ChatGPT Work.
| Use this | When you want |
|---|---|
| Chat | An answer, an explanation or a quick draft |
| ChatGPT Work | A one-off or scheduled task, such as research or a file |
| Codex | To build, debug or review software |
| A dot | Something to own an ongoing job and follow up as things change |
Source: OpenAI's comparison table on the Meet dots page.
Who can use OpenAI dots right now
Dots are on three plan types only, and they're rolling out gradually, so an eligible account may not see them on day one. OpenAI's access section lists:
| Plan | Price (USD) | Dots? |
|---|---|---|
| Free, Go ($8), Plus ($20) | $0 to $20 a month | No |
| Pro 100, Pro 200, Pro 500 | $100, $200 or $500 a month | Yes, 18+, outside the EEA, UK and Switzerland |
| Business, Premium seat | $125 a month, or $100 billed annually | Yes, rolling out worldwide |
| Business, Standard seat | $25 a month, or $20 billed annually | No |
| Enterprise, Edu, Healthcare | Contact sales | Beta, off until an admin enables it |
Prices from OpenAI's ChatGPT pricing page and Business pricing page, checked October 10, 2026. Business needs at least two seats.
Your first dot is included at no extra cost. OpenAI's launch post says extra dots and more speed or workload will come "in the future", with no prices yet.
In Britain or Europe, no Pro tier gets you a dot; our OpenAI dots UK and EU guide covers the Business Premium route. Choosing a Pro tier? See our ChatGPT Pro $100 vs $200 vs $500 comparison.
What counts toward your limits
Talking to your dot doesn't use your ChatGPT usage limits. Tasks it starts in ChatGPT Work or Codex do count toward those products' limits as usual. OpenAI says plans include an allowance for "deeper work, with extended limits for the first month after launch", so expect the terms to tighten once that first month ends.
How to use OpenAI dots: set up your dot step by step
Setup takes three steps, and you can now do it from your phone. OpenAI's October 5β9 update added mobile setup, so older guides that say "desktop only" are out of date.
- Open dots. Use the latest ChatGPT mobile app, the desktop app, or ChatGPT in a desktop browser. Mobile web isn't supported.
-
Name it and pick a look. Choose a name, shape, color, eyes, glasses and accessories. Your dot starts with a handle like
@yourname-dot, and naming it changes the handle (a dot named Alfred becomes@yourname-alfred). - Connect plugins, or skip. Add the apps you want it to use, such as email, calendar and files. You can skip this and add them later, which is what we'd suggest (more on that below).
After setup, the same dot follows you across phone, desktop and web.
Optional: connect your computer
You don't need to. The dot has its own cloud computer. If you want it to use local files, your signed-in browser or local Codex tasks, the computers and apps guide says to open the dot's profile in the ChatGPT desktop app, go to Computers, find Your computer and select Allow access, then confirm.
Only one personal computer can be connected at a time, and it has to be online with the ChatGPT app open while the dot uses it. Revoke access removes the connection. An "Offline" status doesn't.
Optional: add Slack or Teams
Open the dot's profile and select Add. In Slack you can DM your dot or mention it in a thread, and by default it only answers you. Adding it to a channel doesn't make it watch that channel; you have to ask. Texting is "coming soon", per the messaging page.
How to lock down a dot's permissions
Each connection is separate: connecting Slack doesn't connect your inbox, and Gmail doesn't connect your computer. So you can give a dot exactly what it needs and nothing more.
OpenAI also runs an automatic review before any action that could affect your accounts or share information, deciding whether the dot can proceed, needs approval, or must hand the step to you (changing a password always goes to you). On top of that, you have three layers of control.
Layer 1: plugin permissions
Your existing ChatGPT app permissions apply to your dot. OpenAI's example is allowing a dot to read email without allowing it to send email. Start every plugin at the narrowest setting that still lets the dot do its job. You can review this from Custom rules via Open Plugins, under Plugin permissions.
Layer 2: custom rules
Go to Settings > Personalization, then Custom rules under Permissions, and select Add. Describe an action and pick one of four behaviors, as listed in OpenAI's controls guide:
| Rule | What it does | Good for |
|---|---|---|
| Take action without asking | Acts with no approval step | Low-risk, reversible jobs, such as updating your own checklist |
| Take action when you say so | Acts only when you explicitly ask; otherwise asks first | Posting a draft you've already reviewed |
| Ask before taking action | Always asks for approval | Emailing clients, booking anything, sharing files |
| Hand off to you | Makes you do it yourself | Deleting shared files, payments, account changes |
A sensible starting set for most people:
- Ask before taking action for sending any email or message to someone other than you.
- Ask before taking action for sharing files or changing who can see them.
- Hand off to you for deleting files, emails or calendar events.
- Hand off to you for anything involving money.
Two caveats from OpenAI's docs: custom rules are "instructions your dot tries to follow, and it can make mistakes," and they can't grant access that a plugin permission blocks. In a work workspace, an admin can switch custom rules off entirely.
Layer 3: plain instructions
For style and process, just tell it. OpenAI notes that asking your dot to draft replies "doesn't give it permission to send them." A line like "Show me every draft before it goes out, and never reply in a channel without checking with me" is worth adding on day one.
Website sign-ins
When a site needs a login, the dot sends a private sign-in form, so your credentials go to its cloud browser outside the chat. Save to Passwords is optional, and reusing a saved login always needs your confirmation.
How to pause, stop or delete a dot
This is where people are most likely to get caught out, because the three controls do different things:
- Pause stops the dot's current main task only. It doesn't stop delegated tasks or cancel scheduled runs. Resume restarts it.
- Activity (in the desktop app, from the dot's profile) is where you open and stop individual delegated tasks.
- Scheduled is where you disable or delete recurring tasks.
Ending a voice call doesn't necessarily stop work, and nothing undoes actions already taken. Delete removes the dot permanently, but it doesn't reverse changes made in connected apps or recall messages already sent.
When you set a recurring task, OpenAI recommends giving a time zone and an end date, then asking the dot to confirm the schedule. For example: "Check the planning channel each weekday at 9 AM Eastern for the next four weeks. Message me only if a deadline is at risk. Confirm the schedule."
Why your dot might not be working
If you can't see dots, or your dot keeps failing, these are the documented reasons:
- Not rolled out yet. OpenAI says you may not see dots immediately even on an eligible plan.
- Wrong plan or region. Plus and Go don't qualify, and Pro is excluded in the UK, EEA and Switzerland.
- Mobile web. It isn't supported. Use the app.
- Enterprise. An admin must enable Use dots (Beta) under Workspace settings > Permissions & roles, per the admin guide.
- A website blocks the cloud browser. Some do. You can ask the dot to try your connected computer instead.
- An expired connection. Open the dot's request and reconnect the plugin.
There are wider teething problems too. The Register reported on October 7 that users were seeing task failures, connection issues and dots that disappeared. OpenAI's October 9 update fixed several bugs, including Outlook setup getting stuck.
Verdict: how to start with a dot
Treat your first week as a trial with a short leash. Connect one or two apps with read-only access, give the dot a single ongoing job that you'd normally check yourself (a project tracker or a weekly inbox summary), and require approval for anything that leaves your account. Review what it does in Activity every day or two. Widen its permissions only once you've seen that its drafts and decisions hold up.
On Plus and tempted to upgrade just for this? We'd wait: the cheapest route is $100 a month and the extended launch limits end after the first month. For more OpenAI changes, see our AI news this week roundup.
FAQ
Can ChatGPT Plus users get dots?
No. OpenAI lists only Pro ($100, $200 or $500 a month), Business Premium seats and Enterprise, Edu and Healthcare workspaces as eligible. OpenAI says it plans to expand dots to more users, but it hasn't said when or to which plans.
How much do OpenAI dots cost?
Your first dot is included at no extra cost with an eligible plan. The cheapest personal route is ChatGPT Pro at $100 a month. A Business Premium seat is $125 a month, or $100 billed annually, and Business needs at least two seats. Extra dots aren't priced yet.
Can I set up a dot on my phone?
Yes, since OpenAI's October 5β9 update. You can create, name and customize a dot and connect plugins in the latest ChatGPT mobile app. Mobile web isn't supported, and connecting your own computer still has to be done from the desktop app.
How do I stop my dot from sending emails?
Set the email plugin to read-only in its plugin permissions, then add a custom rule under Settings > Personalization > Custom rules set to "Ask before taking action" or "Hand off to you" for sending messages. Telling it to draft only also helps, but drafting doesn't grant permission to send.
Does pausing a dot stop everything?
No. Pause stops only the current main task. You have to stop delegated tasks in Activity and disable recurring tasks in Scheduled separately, and stopping work doesn't undo actions already completed.
Why can't I see dots in ChatGPT?
The most common reasons are that the rollout hasn't reached your account yet, your plan isn't eligible, you're on Pro in the UK, EEA or Switzerland, or you're using mobile web. In an Enterprise workspace, an admin has to enable dots first.
Sources checked on October 10, 2026. We'll update this guide when OpenAI changes dots pricing, plans or controls.
This article was first published on TrustLoud, where it's kept up to date as OpenAI changes dots.
Top comments (0)