What are ChatGPT dots? OpenAI's always-on agents, who gets them and how they stay safe

Dots are OpenAI’s new always-on agents, introduced at its DevDay event on September 29, 2026. Where a normal ChatGPT conversation waits for you to type, a dot is meant to keep working after the conversation ends: it watches the apps you connect, runs scheduled checks, researches in the background and comes back with finished work or a question. This guide explains what a dot is, who can use one, how it differs from ChatGPT Work and Codex, and the safeguards OpenAI describes. It draws on OpenAI’s launch post and Help Center, and everything here is OpenAI’s own description. We haven’t tested dots.

What a dot is

OpenAI calls dots “frontier intelligence that have your back”. In practice a dot is a personal agent with four ingredients:

• Its own cloud computer and browser. A dot does its work on a computer of its own, which you can open at any time to inspect what it did.
• GPT-6 Astra. Dots run on OpenAI’s most capable model. See ChatGPT models explained for where Astra sits among the others.
• Apps through plugins. OpenAI says a dot can connect to more than 4,000 apps through its plugin ecosystem.
• Memory and learning. A dot receives memories from ChatGPT, creates its own, including from connected apps, and learns your preferences from feedback over time.

You start with a primary dot, give it a name and a character or pet, and OpenAI says it envisions teams of dots in future. OpenAI’s examples include a developer’s dot watching customer feedback and preparing tested pull requests, a content creator’s dot turning an interview transcript into clips, show notes and draft posts, and an early tester whose dot noticed an unsent invoice, prepared it and sent it after he approved.

Who can use dots and what they cost

Your first dot is included in a Pro or Business Premium plan at no extra cost, with an allowance for deeper work that has extended limits for the first month. OpenAI says you will later be able to add more dots, and to scale each dot’s output by raising its speed or its monthly workload. You must be 18 or older. For the price of Pro itself, now $100, $200 or $500 a month, see ChatGPT Pro explained.

How dots count against your limits

Conversations with your dot don’t count toward your ChatGPT usage limits. If your dot starts or manages tasks in Codex or ChatGPT Work, those tasks count as usual. For the first month after launch, OpenAI says dots usage won’t count toward eligible Pro, Business and Enterprise plan allowances, and it will share usage terms for each plan afterwards.

How to set one up

According to OpenAI’s help page:

1. Create your dot on desktop. Use the ChatGPT desktop app (also available on Windows) or ChatGPT on desktop web. You cannot create a dot on mobile, and dots are not supported on mobile web.
2. Follow the onboarding prompts and name your dot. Its default handle is @yourname-dot, and you can change the name and avatar later from its profile.
3. Connect apps. Review each plugin’s permissions before connecting it. You can connect your personal email account, but at launch a dot cannot have its own standalone email address.
4. Choose channels. After setup you can message your dot in the ChatGPT mobile app. Slack and texting are connected from desktop.
5. Give it a task or a schedule. Ask it to review your calendar each morning, research a topic or remind you of something. Scheduled tasks appear under Scheduled in its profile, alongside In progress and Completed.

Where you can talk to a dot

You can message or call your dot in ChatGPT on desktop, web and mobile, and message it in Slack and Microsoft Teams. Texting is coming, and where it’s available it is a limited beta for Pro users in the US, run through a third-party provider. OpenAI says to be careful sharing sensitive information by text. A dot cannot start a call to you at launch.

What a dot does on its own

A dot can research in the background, review the apps you connected and form memories from them even when you haven’t asked a question, and run the reminders and recurring tasks you’ve scheduled. OpenAI calls the background part “proactive research”. It says proactive research uses connected apps through tools restricted to read-only, so they can’t send messages, change app content or control your browser or computer.

Your dot can also use its own cloud computer, create cloud tasks in Codex environments that you’ve set up (see Codex Cloud explained), and, if you permit it, create Work or Codex tasks on your own computer, use local skills and use your local browser when its cloud browser is blocked.

Safety and controls

OpenAI lists several safeguards, and says dots “can still make mistakes”:

• Separate computers. A dot works on its own cloud computer, while your computer and its contents stay separate unless you choose to connect it. Local access starts off.
• Saved passwords stay hidden. When signing in to supported websites, dots can use saved passwords without exposing them to the model.
• Custom Rules. For actions such as sharing, purchasing or accessing things, you can choose for each rule whether the dot takes action without asking, acts only if pre-approved, asks first or hands off to you.
• Auto-review. Dots check actions that could affect your accounts or share information against your instructions, rules and OpenAI’s safety requirements, which decides what proceeds, what needs approval and what you must do yourself. Sensitive tasks such as changing a password always stay with you.
• Activity View, pause and reset. You can follow progress, including background work, redirect the dot, pause it or reset it. Reset deletes its conversations, saved memories and scheduled tasks.
• Monitoring. OpenAI says its monitoring system can pause or stop a dot’s work if it detects a safety concern, and that it has built defences against malicious instructions.

Privacy

On personal ChatGPT plans you can control whether dots’ conversations and work are used to improve models. OpenAI says it doesn’t train directly on proactive research or a dot’s notes to itself, although information from them may be used if it helps inform an eligible conversation, depending on your settings. It doesn’t use content from Business, Enterprise or Edu workspaces to improve models by default. Disconnecting an app doesn’t delete what your dot has already learned from it, so use Reset for that.

Specialist dots for organisations

OpenAI also previewed specialist dots, which take on dedicated responsibilities inside a company with their own identity, credentials and access, using IT-provisioned hardware and deep integrations with systems of record. It is starting with focused enterprise pilots built on lessons from procurement, invoice processing, email marketing, customer support and commercial contracting, and is working with Microsoft to bring specialist dots into Agent 365 governance controls.

How dots compare with other agents

Dots sit alongside several similar ideas. ChatGPT Work does longer tasks you start and supervise, while a dot is designed to keep going between conversations. Google’s Gemini Spark and Anthropic’s Claude Cowork take different routes to the same goal, and we compare all three in dots vs Gemini Spark vs Claude Cowork. For the wider category, see AI agents explained and the rest of the week’s announcements in OpenAI DevDay 2026: everything announced.