OpenAI Dots Explained: Always-On AI Agents
OpenAI Dots are persistent AI agents with connected apps and a cloud computer of their own. This guide explains what always-on means, what a dot can access, how approvals and rules work, who gets Dots and what a dot costs.
Last updated: October 2, 2026
OpenAI Dots are persistent AI agents that run on a cloud computer of their own and keep working on a user's goals after the user closes ChatGPT. OpenAI announced Dots at DevDay on September 29, 2026. Each dot runs on GPT-6 Astra, connects to more than 4,000 apps through plugins, and takes messages in ChatGPT, Slack and Microsoft Teams. The first dot is included with ChatGPT Pro and Business Premium.
OpenAI writes the product name in lowercase, as "dots". This guide covers what always-on means in practice, what a dot can access, how users control a dot, who gets access, the price and the limits. OpenAI's own description is "persistent agents with connected apps and their own cloud computer", from OpenAI's DevDay 2026 announcements post. The announcement is Introducing dots. Setup, access and control details come from OpenAI's "Getting started with your dot" help article. Points taken from launch reports name the outlet.
What are OpenAI Dots?
OpenAI Dots are agents that hold ongoing responsibilities for one user. A standard ChatGPT conversation waits for a message and ends when the task ends. A dot receives a goal, keeps the goal, and returns to the goal without a new prompt.
A dot has 4 parts.
- A model. Dots run on GPT-6 Astra, OpenAI's largest GPT-6 model.
- A cloud computer. Each dot has a dedicated cloud computer and browser.
- Connected apps. A dot reaches more than 4,000 apps through ChatGPT plugins.
- Rules. Built-in rules and user-written Custom Rules decide what a dot does alone and what needs approval.
One dot is included per account today, and OpenAI says users will be able to add more dots later.
What does "always-on" mean in practice?
"Always-on" means the dot runs on OpenAI's cloud, not on the user's device, so work continues when the laptop is closed and ChatGPT is not open. There are 3 practical effects.
- Work continues between conversations. A dot pursues an assigned goal around the clock and adjusts after feedback.
- Research happens in the background. A dot that has no active task does proactive research across connected apps. OpenAI states that background research uses connected apps "with tools that are restricted to be read-only", so the dot cannot send a message or change content during that time.
- Several projects run in parallel. A dot works on more than one project at the same time.
"Always-on" does not mean the dot watches a personal computer. The dot works from the cloud computer OpenAI assigns, and access to the user's local computer is optional and off at the start, per the help article.
What can a dot access and do?
A dot can access the apps a user connects and the cloud computer OpenAI assigns to the dot. A dot has no access to an app the user has not connected.
A dot carries out 4 kinds of work.
- App tasks through plugins, including reading email threads, drafting documents and updating records in connected apps.
- Browser tasks on the dot's own browser.
- Research and analysis started in ChatGPT Work.
- Software tasks started in Codex.
The Next Web quotes one early user, developer Dan McAteer: "My dot proactively picked up on an invoice I needed to send for my freelance writing, pulled the relevant details from the email thread and drafted the invoice." The example shows the pattern: the dot found the task, gathered the details and prepared a draft for the user to review.
How do users control a dot?
Users control a dot through 7 controls: built-in rules, Custom Rules, auto-review, the activity view, credential handling, pause and reset.
| Control | What the control does |
|---|---|
| Built-in rules | Decide when a dot acts alone and when a dot asks for approval |
| Custom Rules | Let the user allow, block or require approval for a specific action |
| Auto-review | Checks actions that affect accounts or share information |
| Activity view | Shows current and background work so the user can redirect the dot |
| Credential handling | Signs in with saved passwords without exposing them to the model |
| Pause | Lets the user pause and resume the dot; OpenAI's monitoring can pause or stop a dot's work too |
| Reset | Deletes the dot's conversations, memories and scheduled tasks |
OpenAI's wording on the default behavior is: "Dots start with built-in rules on when to act alone and when to ask for approval." A Custom Rule takes 1 of 4 behaviors, per the help article: "Take action without asking", "Take action if pre-approved", "Ask before taking action" and "Hand off to you". OpenAI states that "certain sensitive tasks, such as changing a password, always stay with you."
OpenAI adds a caution in the announcement: "Dots can still make mistakes, so always review consequential work." The help article goes further: "Your dot can make mistakes, including when following your rules."
Does a dot have unrestricted access to accounts?
No, a dot does not have unrestricted access to a user's accounts. A dot acts with granted permissions, and 3 boundaries apply.
- Scope. A dot reaches only the apps the user connects through plugins.
- Approval. Rules require the user's approval for defined actions, and Custom Rules can block an action outright.
- Mode. Background research is read-only.
The difference matters when deciding what to connect. A dot with access to email can read and draft email inside the rules the user sets. That is granted access with limits, and the limits are only as tight as the rules. Review the built-in rules and add Custom Rules for sending messages, payments and file sharing before connecting email, files or photos.
Meta sells a competing personal agent with a different permission model.
Who can use OpenAI Dots?
OpenAI Dots are rolling out to ChatGPT Pro and Business Premium users in eligible markets, with a beta for Enterprise, Edu and Healthcare workspaces.
| Plan | Access on October 2, 2026 | Region note |
|---|---|---|
| Pro | Rolling out | Not available in the EEA, Switzerland or the UK |
| Business Premium | Rolling out | All supported ChatGPT regions |
| Enterprise, Edu, Healthcare | Beta, off by default | Workspace admin turns the beta on |
| Free, Go, Plus | No access announced | Not published |
EEA stands for European Economic Area and UK for United Kingdom. The plan and region details come from OpenAI's help article. The rollout is gradual, and OpenAI says access can take several days to reach an eligible account.
How much do OpenAI Dots cost?
The first dot costs nothing extra on Pro and Business Premium. There is no separate dot subscription as of October 2, 2026.
There are 3 cost details to know.
- Dot conversations do not count toward ChatGPT usage limits. Talking to a dot does not reduce the plan's message allowance.
- Delegated work does count. Tasks a dot starts in Codex or ChatGPT Work draw on the plan's allowance.
- Paid add-ons are planned. OpenAI says users will later pay for more dots or for a faster dot, per The Next Web. No price is published.
OpenAI launched ChatGPT Pro 500 on the same day at $500 per month with 25 times the Plus usage allowance. A larger allowance gives delegated dot work more room.
What are specialist dots?
Specialist dots are role-specific agents for companies, in enterprise preview. A specialist dot has an identity, credentials and access to company systems of its own. The 5 roles OpenAI named are procurement, invoice processing, email marketing, customer support and commercial contracting, per The Next Web. OpenAI is working with Microsoft on an Agent 365 integration for managing specialist dots.
How do you start using a dot?
To start using a dot, open ChatGPT in the desktop app or on desktop web with an eligible Pro or Business Premium account. OpenAI's help article sets 4 conditions for setup.
- Creation is desktop only. A dot is created in the Windows or Mac desktop app or on desktop web, not in the mobile app or on mobile web.
- Mobile works after creation. The ChatGPT mobile app reaches the dot once the dot exists.
- Local computer access is optional. Access to the user's own computer is off by default.
- Email is a connection, not an address. A personal email account can be connected, and a dot cannot have an email address of its own.
Beyond those conditions, 3 decisions shape what the dot does. These 3 are our suggestions, not OpenAI's steps.
- Choose the apps to connect. Connect the smallest set of apps the first goal needs.
- Review the rules. Read the built-in rules, then add Custom Rules for actions that send, pay, share or delete.
- Assign one clear goal. Start with one responsibility, check the activity view, then add a second goal.
Dots connect to apps through plugins, the same system that replaces Custom GPTs on December 11, 2026.
What are the limits of OpenAI Dots?
OpenAI Dots have 5 limits as of October 2, 2026.
- Plan limit. Free, Go and Plus accounts have no announced access.
- Region limit. Pro accounts in the EEA, Switzerland and the UK are excluded.
- Quantity limit. One dot is included per account.
- Channel limit. Texting is a limited beta for Pro users in the United States, and a dot cannot start a voice call.
- Accuracy limit. OpenAI states that dots make mistakes and that users must review consequential work.
On data use, OpenAI states that users on personal ChatGPT plans control whether their content is used for model improvement. OpenAI has not published a retention period for dot data or the cloud computer's specifications on the pages reviewed for this guide, and Reset is the documented way to delete a dot's data. All other DevDay announcements are in our OpenAI DevDay 2026 recap.
Frequently asked questions
What is a dot in ChatGPT?
A dot in ChatGPT is a persistent agent with connected apps and a cloud computer of its own that works on a user's goals continuously.
Are OpenAI Dots available on ChatGPT Plus?
No, OpenAI has not announced Dots for ChatGPT Plus. Dots are rolling out to Pro and Business Premium, with a beta for Enterprise, Edu and Healthcare.
Can a dot send emails without asking?
A dot sends an email without asking only where the rules allow it. Built-in rules set when a dot asks for approval, background research is read-only, and a Custom Rule can require approval for every sent message.
Which model do OpenAI Dots use?
OpenAI Dots use GPT-6 Astra, per OpenAI's announcement.
Are OpenAI Dots available in Europe?
No, Pro users in the European Economic Area, Switzerland and the United Kingdom do not have Dots yet. Business Premium users get Dots in all supported ChatGPT regions, per OpenAI's help article.