How to create your first dot, step by step

To create your first dot: confirm your plan has access, open ChatGPT in the desktop app or in your computer’s browser, create the dot, name it, connect the apps you want it to use, and review the approval rules before delegating real work. After initial setup you can message it from your phone.

Before you start

Three preconditions, in order:

1. A plan with access. dots begins reaching Pro and Business Premium. On Enterprise, Edu or Healthcare, a workspace admin has to enable the beta.
2. An enabled market. Availability is by country, not by language. Check the enabled markets list.
3. A computer to start on. Setup happens in the ChatGPT desktop app — there is a Windows version — or on desktop web. You cannot create a dot on mobile, and mobile web is not supported. Once it exists, you can talk to it from your phone.

Step 1: create the dot

Open ChatGPT in the desktop app or your computer’s browser and create your first dot. OpenAI describes the experience as letting the dot introduce itself after initial setup.

Step 2: name it and customize it

You can give it a name and customize it at creation. That is not cosmetic: a name makes it easier to tell dots apart once you have more than one, and helps you treat the dot as a responsibility rather than another tab.

How naming works:

• The default handle is @yourname-dot.
• Once you name it, it becomes @yourname-agentname.
• To change it later, open the dot’s profile and use the pencil icon.

You can also pick a character or a pet from the available options. OpenAI notes the dot may even generate a pet for you.

One logistics detail: messaging channels — Slack and text — are set up from the desktop app, not from mobile.

Step 3: connect apps deliberately

This decision has the biggest effect on how useful your dot turns out to be. The plugin ecosystem covers more than 4,000 apps, but start with few:

• Pick the apps where the work you want to delegate actually lives.
• Start with the ones carrying the least write risk.
• Remember that in proactive research the dot can only read: it will not send messages or change content on its own.

Step 4: define approval rules

dots ship with built-in rules defining when they may act alone and when they must ask for approval. On top of those you can add your own to allow, require approval for, or block specific actions.

A reasonable starting point: leave reads open, require approval for anything that leaves your account (messages, posts, sends), and block anything touching credentials or payments.

Step 5: first assignment

Start with something small and checkable. Do not ask it to run your business: give it an objective with a result you can inspect, and watch how it works. The dot’s computer can be opened at any time to see exactly what it did.

Step 6: review and adjust

After the first assignment, review the activity view. That is where you see progress, including background work, and where you can give new direction. It is also where you notice that permissions need tightening.

Common mistakes

Mistake	Why it hurts
Connecting every app up front	Widens the permission surface for no reason and makes auditing harder
Skipping approval rules	Everything runs on default judgement instead of yours
Asking for one huge, vague objective	There is no way to tell whether the result is good
Never opening the dot’s computer	You lose the main auditing benefit of the design

What comes next

Once the first dot works, the natural next step is thinking in responsibilities: instead of a task list, give it a concrete area. If your organization is heading that way, specialized dots — with their own identity and credentials — are the path OpenAI is opening first at enterprise level.