Its own computer
Each dot has a cloud computer and browser. You can open it any time to watch or take over. Your laptop stays separate unless you connect it.
A dot is an always-on AI agent inside ChatGPT. It has its own cloud computer, connects to the apps you choose, and keeps working on your goals between conversations. It learns how you like things done and comes back with finished work for you to approve.
A normal ChatGPT chat stops when you close it. A dot keeps going: it can research, build, follow up and run scheduled jobs on its own computer. It still asks you before anything important happens.
Each dot has a cloud computer and browser. You can open it any time to watch or take over. Your laptop stays separate unless you connect it.
Give it a goal and it keeps making progress. It runs scheduled jobs, and while you're away it reads your connected apps (read-only) to find ways to help.
Message it or call it in ChatGPT on desktop, web and mobile, or in Slack and Microsoft Teams. It keeps the same context everywhere.
You decide what it may do alone and what needs your approval. Risky actions get an automatic safety review, and you can pause it any time.
You can only create a dot on a computer. Once it exists, you can talk to it from your phone. Tap any step to open it.
Open ChatGPT, click your name, then Settings → Account, and confirm you're on Pro or Business Premium. On a company Enterprise workspace, ask your admin to turn on the dots beta.
Use the ChatGPT desktop app (Mac or Windows) or chatgpt.com in a desktop browser. You can't create a dot on mobile or in a phone browser.
Open the dots page and click Create your dot.
The setup asks a few questions, then your dot introduces itself. It already starts with what ChatGPT memory knows about you. Tell it your role, your goals for the next 90 days, and what "good work" looks like to you. This one conversation shapes everything it does next.
Everything lives in your dot's profile: click its name or avatar.
Name it during setup or later. Its handle changes from @yourname-dot to @yourname-name. Pick a character or a pet as its avatar; your dot might even generate a pet for you.
Describe an action, like "sending emails", "buying anything" or "posting in shared channels", and choose how your dot handles it:
Talk to it like a new hire: your clients, your tone of voice, your priorities, the people who matter. Correct it when it's off. It learns from your feedback, and what ChatGPT remembers about you is under:
Connect messaging channels from the desktop app, then mention your dot in a channel so it can pick up questions and next steps there. Texting is a limited beta for US Pro users.
If it needs files or apps that live on your laptop, open the ChatGPT desktop app on that computer, find Computers → Your computer, then select Allow access and confirm. The computer has to stay online with the app open. Revoke access the same way.
A plugin is how your dot reaches Gmail, your calendar, Notion, your CRM and 4,000+ other apps. Your dot uses the same plugins as the rest of ChatGPT, with the same permissions. A dot with no plugins can only research the web.
Click Plugins in the ChatGPT sidebar (web, desktop or mobile). From your dot's profile, the same list is under Customize.
Browse Popular or New & Noteworthy, or search by name. Open a plugin to see its apps, skills and setup requirements. A Desktop only label means it won't run on the web.
Click Install plugin, then Connect, then sign in to the provider. Double-check it's the right account, for example work Gmail rather than personal.
Being able to read something isn't the same as being able to change it or send messages. Start read-only for a few days, then allow drafts, and only allow full actions once the drafts are consistently right.
Ask your dot something only the plugin can answer: "What's on my calendar tomorrow, and which emails from this week still need my reply?" If it can't answer, reconnect the plugin or check its permissions.
Triage, reply drafts, meeting prep, clash alerts.
Keep plans, reports and decision logs current.
Summaries, weekly updates, unanswered questions.
Lead replies, pipeline hygiene, proposals.
Repurpose, design, schedule. Paid tools use credits.
Shortlists and fare watches. You approve every booking.
Tell us a little about your work, either with the quick form or by talking to the planner. You'll get ready-to-paste automations: which plugins to connect, the exact instruction to give your dot, when it should run, and the Custom Rule that keeps it safe.
No. Conversations with your dot don't count toward your ChatGPT usage limits. Tasks it starts in Codex or ChatGPT Work count as usual. Your plan also includes an allowance for deeper work, with extended limits in the first month after launch.
Only if you let it. Actions that could affect your accounts or share information go through an automatic review against your instructions and Custom Rules. Many need your approval, and some, like changing a password, always stay with you.
You can call your dot from its conversation. At launch it can't start a call to you. It reaches you with messages in ChatGPT, Slack or Teams. Texting is a limited beta for US Pro users.
Not at launch. You can connect your own email account through a plugin so it can work on your behalf.
On personal plans you control whether dot conversations are used to improve OpenAI's models. Business, Enterprise and Edu content isn't used for training by default. OpenAI says it doesn't train directly on proactive research or your dot's private notes.
Yes. Like any AI, it can get things wrong, including when following your rules. Keep each job narrow, start read-only, and always review anything consequential.