ChatGPT dots: what they are and how to use them

What ChatGPT dots are

OpenAI calls dots “remarkably capable, always-on agents built to handle everything.” A dot runs on GPT-6 Astra and works on its own cloud computer with its own browser, using the apps you connect. You talk to it in ChatGPT on desktop, web or mobile, in Slack and Teams, by voice in the mobile app, and in a US beta for Pro, by text message.

• It works while you don't: with read-only access to your connected apps, it looks for ways to help, and it runs scheduled reminders and recurring checks.
• It runs several projects at once and checks in with progress, questions and decisions that need you.
• It remembers: it starts with your ChatGPT memories and learns your preferences as you work together.
• It writes code by delegating: a dot can start Codex tasks, which count toward your Codex limits.
• You can watch it: open the dot's computer at any time, take over, and hand control back.

Who can use dots, and what they cost

Your first dot is included in the plans below at no extra cost. During the first month after launch, dots usage didn't count toward plan allowances; OpenAI says it will share per-plan usage terms after that, and hadn't published them when we checked. Buying more dots or more work per dot is announced, without prices yet.

How to set up your first dot

You need to be 18 or older. The rollout is gradual, so access can take a few days to reach your account.

1. Open dots

   Go to chatgpt.com/dots, or open dots in the ChatGPT desktop app, on desktop web, or in the mobile app on iOS or Android.

2. Create and name your dot

   Follow the onboarding: give it a name, which becomes its handle (@yourname-agentname), and pick an avatar. You can change both later from its profile.

3. Connect your apps

   In the dot's profile, open Customize, then Plugins, choose an app and review the permissions it asks for. Connect Slack from desktop. A dot can use a personal email account you connect, but can't have its own email address yet.

4. Set custom rules

   Under Customize, Custom rules, choose for each kind of action whether the dot should take action without asking, take action only if pre-approved in your prompt, ask before taking action, or hand off to you.

5. Give it a goal

   Describe the outcome you want and share context; attach files with +, and reply in the thread to steer it. Ask for reminders or recurring checks, and manage them under Scheduled.

6. Follow, pause or delete it

   The profile shows work In progress, Scheduled and Completed. Pause the dot from the ••• menu when you need to. Deleting it removes its memories and scheduled tasks and can't be undone.

Source: OpenAI's announcement of dots

Safety: what a dot will and won't do on its own

• Changing a password or sending money always requires you to take over and do it yourself.
• Deleting data permanently or installing software can need your approval each time.
• Auto-review checks risky actions, such as who an email goes to, against your instructions and rules, and your custom rules can't switch it off.
• A dot can buy with a card saved on a merchant's site, but only with your approval each time or a specific approval in advance.
• Your own computer stays separate unless you connect it with Allow access in the ChatGPT desktop app; it's off by default.
• OpenAI says its prompt-injection protections reduce the risk but don't eliminate it, and that some completed actions can't be undone.

What dots don't do yet

• You get one primary dot; teams of dots are described as a future direction.
• A dot has no email address of its own, can't call you, and doesn't work on mobile web.
• There are no webhooks or event triggers to start it; it works from your messages, its schedules and its own read-only research.
• You can't view, edit or delete individual memories; deleting the dot is the only way to clear what it learned.
• Plus users don't get dots, and Pro users in the EEA, Switzerland and the UK don't either.

ChatGPT dots vs Grok Bot vs Crewmark

xAI announced Grok Bot in August 2026 as “your team of always-on agents” with their own computer. Crewmark takes a different route: instead of a new agent in someone's cloud, it turns the coding agent you already use into an employee on your own machine.

Already pay for Claude Code or Codex? Make it always-on

If your always-on work is in a code base or a website, you don't need a new agent. Crewmark gives the Claude Code or Codex session you already use a role, a goal and working hours, and wakes it on a heartbeat to take the next issue from a board.

1. Turn a session into an employee

   Pick a role such as SEO specialist or software engineer, write its goal as a number, and run the setup command in Claude Code or Codex.

2. Let the heartbeat keep it working

   During its working hours the employee wakes every 15, 30 or 60 minutes in the same conversation, so it remembers what it did; see how to make Claude Code proactive.

3. Assign issues and answer its messages

   Put work on the board; the employee picks it up, comments as it goes and messages you when it's blocked. Read what an AI employee is.

ChatGPT dots FAQ