A ChatGPT dot is an always-on assistant that keeps working on a job after you close the chat, and you can set one up in about ten minutes without writing any code. You create it in the ChatGPT desktop app, a desktop browser or the mobile app, give it a name, connect only the apps it actually needs, and hand it one ongoing responsibility, such as “keep this project on track and tell me when a deadline is at risk.” This guide walks through each step in OpenAI’s own documentation, explains what each connection really allows, and covers the settings that keep it from doing things you didn’t mean to approve.

Key facts (from OpenAI’s announcement, Help Center and ChatGPT Learn docs):

• What it is: an agent in ChatGPT, powered by GPT-6 Astra, with its own cloud computer and browser. It keeps working between conversations and messages you when it has results or needs a decision (OpenAI).
• Who gets it: Pro 100, Pro 200 and Pro 500 users over 18 outside the European Economic Area, the UK and Switzerland; Business Premium worldwide; Enterprise (including Edu and Healthcare) only after a workspace admin turns it on (ChatGPT Learn). It is rolling out gradually, so access can take several days to reach your account.
• Cost: your first dot is included in Pro and Business Premium at no extra cost. Talking to your dot doesn’t count toward ChatGPT usage limits, but tasks it starts in ChatGPT Work or Codex count toward those products’ limits as usual (OpenAI).
• Where you can set it up: ChatGPT desktop app (Mac and Windows), a desktop browser, or the latest ChatGPT mobile app. Mobile web is not supported (OpenAI Help Center).
• Where you can talk to it: ChatGPT (text or voice call), Slack and Microsoft Teams. Texting is listed as “coming soon” (ChatGPT Learn).

Is a dot the right tool for your job?

OpenAI now has several ways to get work done inside ChatGPT, and dots overlap with them. OpenAI’s own comparison is a good rule of thumb (ChatGPT Learn):

Use this	When you want
Regular chat	An answer, an explanation, ideas, or a draft in the conversation
ChatGPT Work	To hand off one task (research, analysis, a file), now or on a schedule
Codex	To build, debug, test or review software
Your dot	Someone to keep track of an ongoing responsibility, work out next steps as things change, and follow up between conversations. It can delegate pieces to ChatGPT Work or Codex

So a dot is not a better chat box. It earns its keep on jobs that drag on for days or weeks and need someone to keep checking: an event you’re organizing, a proposal that keeps changing, a content pipeline, or a stream of feedback. If you only need one answer, a normal chat is faster.

Step 1: Check that you have access

1. Make sure you’re on an eligible plan (see the key facts above). On Pro, you also need to be over 18 and outside the EEA, UK and Switzerland.
2. Update the ChatGPT app. OpenAI asks for the latest mobile app version for setup on a phone; mobile setup itself only arrived in the week of October 5–9, 2026 (ChatGPT Learn, What’s new).
3. Look for dots in ChatGPT on the desktop app or desktop web. If you don’t see it, it may simply not have reached your account yet. OpenAI says the rollout is gradual and you may not see dots right away even if your plan qualifies.

On a company plan: if you’re on ChatGPT Enterprise and see nothing, your admin has to switch it on. The setting is Workspace settings > Permissions & roles > Permissions > Workspace default > Workspace capabilities > Use dots (Beta), and it’s off by default (OpenAI Help Center). Admins also control separate switches for Slack and Teams, local computer access, custom rules, and the cloud browser, cloud network and password manager.

Step 2: Create and name your dot

OpenAI’s setup is three steps (ChatGPT Learn):

1. Open dots in the ChatGPT desktop app, a desktop browser or the mobile app, and follow the introduction.
2. Give your dot a name and pick how it looks. You can choose its shape, color, eyes, glasses and accessories, or pick a pet. You can keep the defaults and change them later.
3. Connect plugins for the apps you want it to use, such as email, calendar and files. You can skip this and add them later, and for a first run that’s the better choice (see Step 3).

Your dot gets a handle. It starts as @yourname-dot, and naming it changes it to @yourname-agentname. OpenAI’s example is @tibo-alfred for a dot named Alfred. To rename it or change the avatar later, open its profile by selecting its name or avatar, then select the pencil icon (OpenAI Help Center).

After setup you can keep using the same dot on any of your devices. On the phone, you can also tell ChatGPT to open straight to your dot’s conversation.

Step 3: Decide what to connect (and what to leave off)

This is the step that matters most for safety, and OpenAI’s docs are clear that the three kinds of connection are separate (ChatGPT Learn):

Connection	What it lets your dot do
A messaging channel (Slack, Teams)	Talk to you and send updates there. It does not give access to your inbox, other apps or your computer
A connected app or plugin (Gmail, Google Drive, GitHub…)	Use that service’s information and tools, within the permissions you set for it
Your local computer	Work with that computer’s files, apps and tasks while it’s online

A few practical rules follow from this:

• Start with the apps the first job needs, nothing more. If you want help with an offsite, that might be your email and calendar. Leave everything else off until you’ve seen how it works.
• Your existing ChatGPT app permissions carry over. OpenAI’s example: you can let it read email without letting it send email (ChatGPT Learn). Review each plugin’s connected account and allowed actions. In custom rules (Step 6) there’s an Open Plugins button that takes you to Plugin permissions.
• Your dot reads connected apps on its own. Connected apps can be reviewed “proactively,” and your dot can form memories from them even when you didn’t ask about them (OpenAI Help Center). OpenAI says that background research is read-only: it can’t send messages, change app content or control a browser. But it does mean anything in a connected inbox may end up in your dot’s notes, so connect accounts with that in mind.
• Your computer is optional and off by default. Your dot already has its own cloud computer, so most people don’t need to connect a personal one at first. We cover it in Step 8.

Step 4: Give it its first responsibility

OpenAI’s advice is to start with “work you want your dot to keep track of, rather than treating each conversation as a fresh request” (ChatGPT Learn). A good first prompt has four parts:

1. The goal. What should stay on track?
2. The sources. Which emails, files or channels it should use. You can attach a file or photo with the + button in the conversation.
3. What deserves your attention. Budget changes? Deadlines this week?
4. What it must not do on its own. For example, “draft replies but don’t send them.”

OpenAI’s own example, which is a good template to adapt:

Help me keep our offsite on track. Read the plan and venue emails I shared, and make a list of decisions, deadlines, and things we’re waiting on. Focus on anything that affects the budget or needs an answer this week. Draft suggested replies for my review, but don’t send them.

When the first result comes back, correct what’s missing before you widen the job. OpenAI recommends exactly that: review the first result and fix the details before expanding the responsibility.

When you first open it, your dot introduces itself and uses ChatGPT memory and any connected apps to suggest where it could help. You can start talking while it gathers context. Its suggestions are only as good as what you’ve connected, so a dot with no apps connected will mostly ask you questions.

Step 5: Set up check-ins and scheduled work

Your dot decides on its own when to follow up, so you don’t have to schedule everything. But anything that must happen at a fixed time needs a saved schedule, and OpenAI lists four things to include (ChatGPT Learn):

• What to check or update
• When to run, including the time zone and an end date
• Which changes deserve a notification
• Where to send the results

Example from OpenAI’s docs:

Check the connected planning channel each weekday at 9 AM Pacific for the next four weeks. Update the project checklist when a deadline changes. Message me in ChatGPT only if a deadline is at risk or you need a decision. Confirm the schedule.

Always end with “Confirm the schedule.” and read the reply. Then open your dot’s profile and check Scheduled to see the instructions, timing and destination that were actually saved (ChatGPT Learn).

Two related abilities:

• Event-based work. Where a connected service supports it, you can ask your dot to act when something happens, such as a new bug report in a Slack channel. Ask it to confirm which events it can actually follow. Connecting Slack alone doesn’t create a monitoring task.
• Your existing ChatGPT Work automations. Your dot can read and edit them. Name the automation you want changed and ask it to confirm the update.

Step 6: Understand approvals and add custom rules

Before your dot does anything that could affect your accounts or share information, an automatic review checks the action against your instructions, permissions, custom rules and built-in safety requirements. The review decides whether it can go ahead, needs your approval, or must be handed back to you (ChatGPT Learn). According to OpenAI’s privacy and safety FAQ (OpenAI Help Center):

• The most sensitive actions, like changing a password or transferring money, are always handed to you to finish yourself.
• Actions like permanently deleting data or installing software may need your approval every time.
• Some actions, such as sending recurring messages, can be approved in advance.
• Approving one message doesn’t give ongoing permission to contact people for you. For a future message, spell out who it goes to, what it says, and when or under what conditions it’s sent.

The single most useful habit: “draft” is not “send.” OpenAI notes that asking your dot to draft replies doesn’t give it permission to send them. If you want it to send, say so, with a clear scope.

Adding a custom rule

Custom rules are optional. Use them for boundaries you want to hold every time, not for writing style. Where to find them:

• Desktop and web: Settings > Personalization, then Custom rules under Permissions
• iOS and Android: open the sidebar, then Customize > Custom rules

Select Add, describe the action, choose how your dot should handle it, then select Add rule. The four choices (ChatGPT Learn):

Rule	What it means
Take action without asking	Does it without asking for approval
Take action when you say so	Proceeds when you explicitly ask; otherwise asks first
Ask before taking action	Always asks for approval
Hand off to you	Asks you to do it yourself

Sensible starting rules: Ask before taking action for messages to customers or anyone outside your team, and Hand off to you for deleting shared files. Keep in mind what custom rules can’t do: OpenAI says they are instructions your dot tries to follow and it can make mistakes, they don’t grant access to an app or computer, and they can’t remove built-in safeguards or required confirmations such as approving a saved login. On Enterprise, an admin can switch custom rules off entirely, and then your saved rules don’t apply.

Step 7: Message it in Slack or Teams, or call it

You reach the same dot everywhere, and changing channels doesn’t reset its memory (ChatGPT Learn).

• Slack or Teams: open your dot’s profile, select Add and connect the contact method. OpenAI says to use desktop to connect messaging channels. In Slack you can DM it, or add it to a channel and mention it in a thread. By default it only responds to you. You have to tell it to engage with others, and it may ask you privately what you’re comfortable sharing before it replies in a channel.
• Voice: open its conversation and select the phone button, or select Call in its profile on desktop. You can type while you talk. Hanging up ends the call, but assigned work keeps going.
• Tell it where different updates go. For example: routine progress in ChatGPT, decisions in Slack.
• Adding it to a channel doesn’t start monitoring. Say which channels to check, what to look for and when to notify you, and ask it to confirm the recurring task.

iPhone shortcuts: with ChatGPT for iOS 1.2026.279 or later (rolling out), you can add a Your dot Home Screen widget (Edit > Add Widget > ChatGPT > Your dot). In Apple’s Shortcuts app there’s an Open your dot action with Messaging, Dictation and Call modes, which you can even assign to the Action button under Settings > Action Button (ChatGPT Learn).

Step 8: Sign-ins, the cloud computer, and connecting your own computer

Watching and taking over the cloud computer

Open your dot’s profile and look under Computers to see its cloud computer. Opening it only lets you watch. Select Take over to use its mouse and keyboard, then Return control when your dot can carry on (ChatGPT Learn).

When a website needs you to sign in

The cloud browser has its own sessions. Being signed in on your laptop doesn’t sign your dot in. When a site needs a login, your dot sends a sign-in request with a private form. Your credentials go to the remote browser outside the conversation, so don’t paste passwords into the chat. You can also pick Take over to sign in and type them on the site yourself. Save to Passwords is optional, and reusing a saved login for a new sign-in always needs your confirmation. OpenAI also says your dot’s saved context doesn’t keep credentials, images or screenshots (OpenAI Help Center).

If you’d rather keep passwords in a dedicated manager, 1Password listed ChatGPT dots among its upcoming agent integrations; at the time of our 1Password for Claude guide it was still marked “coming soon.” For a comparable approach in another assistant, see how Grok Bot’s vault handles logins.

Connecting your own computer (optional)

Only do this if your dot needs files or apps that live on your machine, or a site blocks the cloud browser. The steps (ChatGPT Learn):

1. Open your dot’s profile in the ChatGPT desktop app on that computer.
2. Under Computers, find Your computer and select Allow access.
3. Review the confirmation and select Allow access again.

Things to know before you do:

• You can connect only one personal computer at a time, and access applies wherever you message that dot, including from your phone.
• The computer must be online with the ChatGPT app open while your dot uses it. Offline means it’s unavailable, not disconnected.
• With access on, your dot can use local files, your signed-in local browser when the cloud browser is blocked, local skills, and Work or Codex tasks on that machine (OpenAI Help Center).
• To remove access, select Revoke access. Connecting the computer to Codex or turning on Work Sync is a separate permission. Neither one alone gives your dot access.

Step 9: Review, pause, stop or delete

Check in on the work regularly. In the desktop app, open your dot’s profile and select Activity; each task shows its progress, files, results, and any request waiting on you, such as a decision, sign-in or approval. The profile also sorts work into In progress, Scheduled and Completed.

Stopping work has three separate controls, and mixing them up is the most common mistake (ChatGPT Learn):

• Pause stops your dot’s current main task only. It does not stop delegated tasks or cancel future scheduled runs. Use Resume to continue.
• To stop a delegated task, open it in Activity and stop it there.
• To end a recurring job, open Scheduled and disable or delete it.

Delete removes the dot and can’t be undone. Save any results first. Deleting it doesn’t undo changes already made in connected apps or recall messages already sent.

Privacy settings worth checking

• Memory is shared with ChatGPT both ways. Your dot can receive ChatGPT memories and recent conversation context, and its conversations can feed ChatGPT memory. Turning off Memory in ChatGPT stops the sharing but doesn’t delete what your dot has already received (OpenAI Help Center).
• Your dot keeps its own notes, separate from ChatGPT’s saved memory, so changing a ChatGPT memory setting doesn’t necessarily change them. Deleting the dot deletes its context.
• Training: your ChatGPT data controls apply to eligible conversations and work. OpenAI says it doesn’t train models directly on your dot’s proactive research or private notes (ChatGPT Learn).
• Private stays private by default. Your dot can use a deadline you told it privately to help you write an update, but sharing private details in a team channel still needs your permission.
• Turn on two-factor sign-in. A dot can act across your connected accounts, so protecting your ChatGPT login matters more. OpenAI points to Settings > Security and login to set up multi-factor authentication.

Troubleshooting

I don’t see dots at all. Check your plan and region first. Pro in the EEA, UK and Switzerland isn’t included. Then update the app and try desktop web. The rollout is gradual. On Enterprise, ask your admin to enable Use dots (Beta).

It won’t open on my phone’s browser. Mobile web isn’t supported. Use the ChatGPT app.

It can’t get into a website I’m logged into. The cloud browser doesn’t inherit your logins, VPN or device policies (OpenAI Help Center). Approve its sign-in request or use Take over to sign in. If the site blocks cloud browsers, ask your dot to try your connected computer.

It drafted something but never sent it. That’s by design: drafting doesn’t imply permission to send. Tell it explicitly who to send to and what, or add a custom rule.

I paused it but it still did something. Pause only stops the main task. Check Activity for delegated tasks and Scheduled for recurring runs.

A connected app stopped working. If a connection expires, your dot raises a request. Open it and reconnect before it can continue.

A task says completed, but the result isn’t there. OpenAI warns that a completed run doesn’t by itself confirm the result was achieved or delivered. Open the task in Activity and read the output and any errors.

My Slack channel is quiet. Adding your dot to a channel doesn’t start monitoring. Tell it what to watch and ask it to confirm the task.

FAQ

Does a dot cost extra? Your first dot is included in Pro and Business Premium. OpenAI says the plan includes an allowance for deeper work, with extended limits for the first month after the September 29, 2026 launch, and that it will share per-plan usage terms after that period. Conversations with the dot don’t count toward ChatGPT usage limits, but Codex or ChatGPT Work tasks it starts do. If you’re deciding between Pro tiers, see our breakdown of Pro 100, 200 and 500 after the October 29 change.

Can I have more than one dot? Not yet. OpenAI says you’ll be able to add more dots in the future. One dot can handle more than one ongoing responsibility.

Which model does it use? OpenAI says dots are powered by GPT-6 Astra. For background on the GPT-6 family and how models are picked in Codex, see our GPT-6.1 Sol guide.

Does it need my computer to be on? No. It works on its own cloud computer, even while your devices are off. Only work on a connected personal computer needs that machine online with ChatGPT open.

Can it do coding work? Yes, through Codex. For cloud coding it needs a Codex cloud environment that you’ve already set up with the repository. You can just ask your dot to start or check a Codex task from your phone.

Is it safe to connect my email? It’s built for that, with automatic action review, approval for sensitive steps, and handoffs for passwords and money. Still, OpenAI says dots can make mistakes, so review consequential work. Start with read-only access where the plugin allows it and widen it once you trust the results. If you use agents in coding tools too, the same “least access first” idea applies to Claude Code’s permission modes.

Sources

• OpenAI: Introducing dots (September 29, 2026)
• OpenAI Help Center: Getting started with your dot, Dots privacy, security, and safety FAQs, Manage dots in ChatGPT workspaces, ChatGPT release notes
• ChatGPT Learn: Meet dots, Get started, Computers and apps, Control your dot, Message your dot, Tasks and memory, What’s new, October 5–9, 2026

For more step-by-step AI guides, see the 7minai homepage.