# Dots in ChatGPT: What They Are, How to Create Your Own Agent, and How to Use It
> How to create a dot in ChatGPT, give it its first task, configure access, check the result, and stop tasks and schedules separately.
Source: https://ailoot.ai/blog/en/chatgpt-dots-create-control/
Canonical: https://ailoot.ai/blog/en/chatgpt-dots-create-control/
Language: en
Publisher: Ailoot (https://ailoot.ai/)
Translation: guide-020
Revision: sha256-05e9feb18968bd90d095ca0dc3121a300b3fa4801de638c58c4d541056cea0e8
Key points
A dot is an always-on agent in ChatGPT powered by the GPT-6 Astra model. It has its own computer and browser in the cloud. It continues working between conversations and returns with results or questions on its own. It can continue working in the cloud even when your computer is turned off. You can create one only in the ChatGPT desktop app or in ChatGPT in a desktop browser. At launch, access is available on the Pro 100, Pro 200, and Pro 500 plans (for users over 18, outside the European Economic Area, the United Kingdom, and Switzerland), Business Premium, and Enterprise. In Enterprise, dots are off by default; a workspace administrator enables them. Access is rolling out gradually: dots may not appear right away even on an eligible plan. Pause stops the current main task: delegated tasks and schedules must be stopped separately.
Learn more about working with models in the [Ailoot documentation](https://ailoot.ai/docs/).
## How does a dot differ from regular ChatGPT?
OpenAI describes a dot as an agent that works with your tools and projects: you can continue the conversation while it carries out an assignment. It brings you results on its own and asks you when it needs your decision.
You can assign a dot an entire area of work. It tracks progress, decides what to do next, and continues when something changes. If you return with new information or priorities, you don't need to start over. It can decide when to pause and when to return to a task, so a fixed schedule isn't needed for every next step. It can also launch background agents to handle several things in parallel while you talk with it.
According to the official documentation, a dot has:
- **The GPT-6 Astra model.**
- **Its own cloud computer and browser.** In the cloud, a dot searches for information, works with files, and runs programs. Its cloud work continues even when your devices are turned off.
- **Its own memory.** A dot starts with work-related information from ChatGPT memory and keeps its own notes about your preferences, decisions, and current work. These notes are separate from saved ChatGPT memory and are not a complete record of everything you've said.
- **Several ways to communicate.** It's the same dot in ChatGPT, Slack, Microsoft Teams, or a call. Switching channels doesn't create a new dot or reset its memory.
Learn more about AI agents in [What Is an AI Agent? A Plain-Language Guide](https://ailoot.ai/blog/en/ai-agents-explained/), and about ChatGPT overall in [ChatGPT: Features, Pricing, and How to Use It](https://ailoot.ai/blog/en/chatgpt-features-plans-getting-started/).
## Who can use dots and what access depends on
The documentation lists no separate price for dots: access depends on your ChatGPT plan. At launch, the terms are:
| Plan | Access to dots |
|----|----|
| Pro 100, Pro 200, Pro 500 | For users over 18 outside the European Economic Area, the United Kingdom, and Switzerland |
| Business Premium | Gradually becoming available |
| Enterprise | Gradually becoming available; off by default and enabled by a workspace administrator |
OpenAI explicitly says that dots may not appear right away, even if your plan is eligible.
Conversations with a dot don't use your regular ChatGPT limits. Tasks that a dot launches or manages in ChatGPT Work or Codex use those products' limits as usual. The plan includes a separate limit for more in-depth work; limits are expanded during the first month after launch.
Compare ChatGPT plans in [ChatGPT Free, Plus, or Pro: Which Plan Should You Choose?](https://ailoot.ai/blog/en/chatgpt-features-plans-getting-started/).
## How to create your own dot
You can create a dot only on a computer: in the ChatGPT desktop app or in ChatGPT in a desktop browser. After setup, you can use the same dot in the mobile app once an update supporting it is released. The mobile web version isn't supported.
To create a dot:
1. Open dots in ChatGPT and complete the introduction.
2. Connect apps: email, calendar, and files. You can skip this step and come back to it later.
3. If you're using the desktop app, decide whether to connect your computer.
After that, the dot will introduce itself and suggest how it can help, based on the information available to it. You can talk to it right away while it learns about your work. It will ask if it needs more information or access.
You can change the dot's name and appearance at any time: its shape, color, eyes, glasses, and accessories. A dot has a short name (handle) in the form `@yourname-dot`. When you give it a name, its handle changes: in OpenAI's example, a dot named Alfred gets a handle ending in `-alfred`.
### What exactly you connect
Communication channels, apps, and your personal computer are connected separately:
- **Communication channel** (Slack, Teams): so you can talk to a dot and get updates from it there.
- **App or plugin** (Gmail, Google Drive, GitHub, and others): so a dot can use that service's data and tools within the permissions you've granted.
- **Your computer**: so a dot can work with its files, programs, and tasks while the computer is available.
Connecting a communication channel in Slack does not give a dot access to your email, other apps, or your computer.
A dot uses the permissions you've already granted ChatGPT to work with apps. For example, you can let it read email without letting it send messages.
To connect your computer:
1. Open the dot's profile in the ChatGPT app on that computer.
2. Under **Computers**, find **Your computer** and click **Allow access**.
3. Read the confirmation and click **Allow access** again.
You can connect only one personal computer at a time. For a dot to use it, the computer must be online and the ChatGPT app must be open. The **Offline** status doesn't revoke access; use the **Revoke access** button to do that.
## How to give a dot its first task
OpenAI recommends starting with work you regularly need to check or update, rather than a one-off question. In your assignment, say what matters to you, which sources to use, and which decisions require your input.
Example of a first assignment based on the documentation:
```
Help me keep track of preparations for the team's off-site meeting. Read the plan and emails from the venues I provided. Make a list of decisions, deadlines, and what we're still waiting for. Prioritize anything that affects the budget or needs a response this week. Draft replies for me to review, but don't send them.
```
According to the documentation, asking for drafts does not give a dot permission to send them. In the example, that prohibition is also stated explicitly.
OpenAI recommends checking the first result and clarifying missing details before expanding the assignment.
### When and where a dot sends updates
Tell a dot where to send updates and when to interrupt you. For example, have it leave progress updates in ChatGPT and send questions that require your decision in Slack. If exact timing matters, ask for a fixed schedule and specify the time zone and duration:
```
Every weekday at 9:00 a.m. my local time (time zone: ) for four weeks, check the connected planning channel. Update my list when a deadline changes. If a deadline is at risk, message me in ChatGPT and prepare a reminder for me to review. Confirm the schedule.
```
For recurring work, the documentation asks you to specify four things: what to check or update, when to run it (including the time zone and, if needed, an end date), which changes to report, and where to deliver the result. Then ask the dot to confirm which schedule it saved.
If the connected service supports it, a dot can respond to an event, such as a new error message in a Slack channel. But adding a dot to a channel does not create a monitoring schedule. You need to say explicitly which channels to watch, what to look for, and when to report it.
## How to limit a dot's actions
Before an action that could affect your accounts or share information with someone, an automatic check compares it with your instructions, permissions, rules, and built-in safety requirements. It determines whether the action can proceed, needs your approval, or requires you to take a step yourself. For example, only you can change your password.
A dot can explore data in connected apps on its own to find ways to help. The tools used for this research can't send messages, change app contents, or control a browser or computer. The same permissions and safety requirements apply to subsequent actions.
### Rules (Custom rules)
You can set optional rules for individual actions. After setup, open **Settings** \> **Personalization**, then select **Custom rules** under **Permissions**. Click **Add**, describe the action, choose an option, and click **Add rule**.
| Option | What the dot does |
|----|----|
| Take action without asking | Takes the action without asking for approval |
| Take action when you say so | Takes the action if you explicitly asked; otherwise, it asks immediately before acting |
| Ask before taking action | Asks for approval before acting |
| Hand off to you | Asks you to do it yourself |
An example from the documentation: choose **Ask before taking action** for sending messages to clients, and **Hand off to you** for deleting shared project files. You can discuss writing style and update format in a conversation; rules determine when a dot can take actions.
Rules are instructions a dot tries to follow, and it can make mistakes. They don't grant access to an app or computer, override built-in safety requirements, or remove mandatory confirmations, such as approval to sign in using a saved login. App permissions are configured separately, under **Plugin permissions**.
### Signing in to websites
When a website requires you to sign in to an account, a dot can send a request using a private form. You enter your username and password in that form, which passes them to the remote browser outside your conversation with the dot. You can also sign in yourself: if you choose to sign in on the website, the dot's browser opens in a mode that you control. Sign in, then return control when the dot can continue working. Simply opening the dot's computer doesn't give you control: you must use **Take over**, then **Return control**.
## How to check a dot's work and stop it
In the desktop app, open the dot's profile and select **Activity**. Select a task to see its progress, files, and results, including background work. If the dot is waiting for a decision, an app connection, a sign-in, or approval, open the request and respond. Schedules are in the **Scheduled** section.
Check the result even after a task is complete. A completed run by itself doesn't confirm that the result was obtained or delivered. Dots can make mistakes, so it's best to check important results and their sources yourself.
The main work, delegated tasks, and scheduled runs are stopped separately:
- **Pause** stops the dot's current main task. Pausing does not necessarily stop all delegated tasks or cancel future scheduled runs.
- Open a delegated task in **Activity** and stop it there.
- Disable or delete a recurring task in **Scheduled**.
- **Resume** continues work after a pause.
Stopping does not undo actions already taken. Deleting a dot (**Delete**) also does not roll back changes in connected apps or recall messages that have already reached people. The confirmation dialog explains what happens to conversations, memory, and schedules when you delete it. Save any results you need before deleting: the action is irreversible.
## Checklist before your first assignment
1. Check that dots are included in your plan and have appeared in your account.
2. Create a dot on a computer.
3. Connect only the apps needed for the first task. Where possible, leave permissions set to read without sending.
4. State explicitly in the assignment that nothing may be sent without your review.
5. Add the **Ask before taking action** rule for sending messages to people.
6. For a schedule, specify the time zone and duration, and ask the dot to confirm.
7. Check the first result in **Activity** and only then expand the assignment.
8. Remember: **Pause**, stopping a task in **Activity**, and disabling a schedule in **Scheduled** are three different actions.
## What we don't know yet
The sources for this article describe dot features and settings; no speed or accuracy measurements of our own are included here. These sources also don't provide a complete list of supported countries: exclusions are named only for Pro. For countries not named among the exclusions, availability is not clear from these sources; check separately whether the feature is available in your account.
A dot works within ChatGPT and follows its rules. If you need an always-on agent on your own server with your own settings, see the separate article: [How to Keep Hermes Agent Running on a Server Around the Clock](https://ailoot.ai/blog/en/hermes-agent-vps-24-7/).
### Sources
- [Meet dots (learn.chatgpt.com)](https://learn.chatgpt.com/docs/dots.md): what dots are, the model, plan access, limits, and connections
- [Get started with your dot (learn.chatgpt.com)](https://learn.chatgpt.com/docs/dots/getting-started.md): creating a dot, the first assignment, and schedules
- [Control your dot (learn.chatgpt.com)](https://learn.chatgpt.com/docs/dots/controls.md): checking actions, rules, stopping, and deleting
- [Connect computers and apps to your dot (learn.chatgpt.com)](https://learn.chatgpt.com/docs/dots/computers-and-apps.md): the cloud computer, signing in to websites, and connecting a computer and apps
- [Tasks and memory (learn.chatgpt.com)](https://learn.chatgpt.com/docs/dots/tasks-and-memory.md): tasks, recurring work, and memory
- [Message your dot (learn.chatgpt.com)](https://learn.chatgpt.com/docs/dots/channels.md): ChatGPT, calls, Slack, and Teams
For your next steps, try [Ailoot](https://ailoot.ai/).
## Questions about Dots in ChatGPT: What They Are, How to Create Your Own Agent, and How to Use It
### Why is a dot still doing work after I press Pause?
Pause stops the main task. Open delegated tasks in Activity and stop them separately, then disable or delete future runs in Scheduled. None of these actions rolls back work already performed.
### Does connecting Slack also connect my email and computer?
No. Communication channels, app permissions, and computer access are separate connections. Grant only the access needed for the first assignment and check the result in Activity before expanding it.
## Ailoot ยท The company and its project portfolio
Ailoot combines established operating experience, a large client base, stable service and ongoing KPI control.
The company and its project portfolio: millions of customers; monthly sales in the billions.
Ailoot brings that operating experience to API access for AI features in customer products, agents, applications and automations.