How to set up your first dot (and fix it when you can't)

Setup takes minutes. Deciding what the dot is allowed to do without asking you takes longer, and that is the part worth doing carefully.

Before you start: check these three things

1. Plan. dots are rolling out to Pro (excluding the EEA, Switzerland, and the UK) and to Business Premium across supported regions. Enterprise access is a beta that a workspace admin must enable.
2. Age. OpenAI states dots are not available to users under 18.
3. Device. You must create your first dot in the ChatGPT desktop app or ChatGPT on desktop web. Creation is not available on mobile.

If any of those fail, skip to the troubleshooting section — you are not doing anything wrong.

Step 1 — Create the dot

In the desktop app or on desktop web, follow the onboarding prompts to create your dot. You can name it, and you can choose from the available characters or select a pet; OpenAI notes a dot may also generate a pet for you.

Name it after the job, not after the personality. "Release coordinator" tells you what to check it on; "Buddy" does not.

Step 2 — Connect apps, deliberately

Connections are managed from the Plugins screen, which uses your shared ChatGPT settings — the same permissions surface used across ChatGPT, ChatGPT work, and Codex.

• Connect one app at a time and read its permissions before confirming.
• Available apps may differ between accounts.
• Start with read-oriented access. A dot's usefulness grows as fast as the permissions you give it, and so does the blast radius of a mistake.

Step 3 — Set the rules that decide when it acts alone

This is the step people skip and later regret. Your dot comes with built-in rules about when it may act autonomously and when it must ask. Custom rules let you go further, per action:

• Allow — the dot may do this without asking.
• Require approval — it must come to you first.
• Forbid — it must never do this.

Two things to know about that system:

• The dot can help you write rules, but changing a rule requires your approval.
• Custom rules operate inside OpenAI's built-in safety requirements. They cannot switch off mandatory confirmations, steps that must be handed back to you, or core safety rules.

Always requires your confirmation: permanently deleting data, installing or running software from unknown sources, granting new security-sensitive access. Always handed back to you: changing passwords, transferring money between financial accounts. Card purchases on merchant sites where you have saved a payment method also need your approval.

Step 4 — Understand what it does when you are not watching

A dot does two kinds of unrequested work:

• Scheduled runs for tasks you set up.
• Proactive research, where it looks for ways to help — OpenAI uses the example of noticing a change to your travel plans.

Proactive research runs in the dot's own cloud environment with read-only tools. By OpenAI's description, those tools cannot send messages, modify content in connected apps, or take control of your browser or computer. Findings are stored as private notes for that dot. Any resulting action still goes through the normal rules and checks.

You manage this in the dot's profile under Scheduled, where tasks show as active, paused, or completed, and where you can change repeat schedule, time, or completion notifications. The desktop app's Activity view shows in-progress and delegated work, and lets you add context, correct a misunderstanding, redirect, or tell the dot to stop.

Step 5 — Add the surfaces you actually want

Once the dot exists, you can message it from the ChatGPT mobile app when mobile access is available, and use Slack or Teams where supported. Mobile web is not supported.

Texting (iMessage/RCS) is a limited beta for Pro users in the US, is not available in Business or Enterprise workspaces, and may not be available to every Pro user. OpenAI notes it uses a third-party provider and warns about sharing sensitive information over text — treat that warning literally.

Troubleshooting: the most common reasons dots are missing

Symptom	Most likely cause
No dots option anywhere	Gradual rollout — OpenAI says DevDay features reach accounts over time; check back
Pro user in the EEA, Switzerland, or UK	Not included in the launch rollout for Pro
Enterprise user, everything else looks right	Beta is off by default; a workspace admin must enable it
Only see it on desktop, not on your phone	Creation is desktop-only; mobile messaging requires mobile access to be live for you
Cannot find chat on mobile web	Mobile web is not supported
Under 18 account	Not available, by policy

Two operational cautions

Secure login is a real safeguard — until you paste a secret. OpenAI states that for supported login flows the model pauses while you fill in the credential form, and that saved-password logins use a dedicated encrypted credential service, so passwords do not enter the model's context. The same documentation notes that if sensitive information is written into a readable message or document, the model can still see it.

Reset is destructive. Resetting a dot deletes it including its conversations, saved memories, and scheduled tasks. If you are experimenting, know which dot you are resetting.

Where to go next

• What dots are, in one page
• Pricing, credits, and what counts against your plan
• Limitations and risks worth knowing before you delegate

Found an error or a fact that has changed since our last update? Tell us — corrections are published, not quietly edited.