What is ChatGPT dots? OpenAI's always-on agents explained
A named agent inside ChatGPT that holds ongoing goals, researches in the background, runs scheduled checks, works in 4,000+ apps and follows rules you set for when to act and when to ask.
LLM Pricing
On this page
ChatGPT dots, also called OpenAI Dots, are always-on agents inside ChatGPT, announced by OpenAI at DevDay on Sep 29, 2026. A dot is powered by GPT-6 Astra, has its own cloud computer and browser, connects to more than 4,000 apps through ChatGPT plugins and keeps working toward your goals between conversations. You reach it in ChatGPT, Slack or Microsoft Teams, and you can talk to it on a voice call.
Facts come from OpenAI's announcement, Help Center and release notes, plus independent guides, linked under Sources.
Short answer
- A dot is a persistent agent with a goal, a memory, a cloud computer and access to your connected apps. It keeps making progress when you are not talking to it.
- Typical uses are scheduled reviews such as a morning calendar brief, background research, follow-ups inside your apps and handing coding work to Codex.
- Custom Rules let you decide, per action, whether a dot acts, acts only when you asked for that exact action, asks first or hands the step to you. A separate reviewer agent checks planned actions before they run.
- Your first dot is included in ChatGPT Pro (from $100.00 a month) and Business Premium ($125.00 per seat). Free, Go and Plus do not include one.
- ChatGPT Pro users in the EEA, Switzerland and the UK cannot get a dot. Business Premium workspaces can.
What a dot is
Earlier ChatGPT agent features ran a task when asked and then stopped. A dot keeps its goal, its computer, its memory and your connected apps between sessions, schedules its own recurring checks and messages you when it needs a decision. OpenAI describes dots as agents that can take on ongoing responsibility.
When you set one up, you give it a name and pick its character. It starts with your existing ChatGPT memories and then forms its own, including from connected apps. App permissions are shared across dots, ChatGPT, ChatGPT Work and Codex, so an app you have already connected is available to the dot.
When it has nothing assigned, a dot does proactive research: it reads the sources you allowed and saves private notes. Those background tools cannot send messages, change content in your apps or drive a browser. A time budget setting guides how long a dot spends on a piece of work.
Each dot runs on GPT-6 Astra, OpenAI's frontier model, which the API lists at $10.00 input and $50.00 output per 1M tokens.
What you can use it for
The documented pattern is a standing responsibility plus a schedule:
| Job | Example |
|---|---|
| Recurring checks | Review your calendar each morning and send a brief |
| Background research | Watch customer feedback or a topic and surface what changed |
| Follow-ups | Notice work that slipped and prepare the next step for approval |
| Coding work | Create tasks in Codex cloud environments, for example to fix an issue it found |
| Purchases | Buy with a card saved on a merchant's site, after your approval |
OpenAI's launch example: a tester's dot noticed that he had forgotten to invoice a publication, prepared the invoice and sent it once he approved. Work in progress, scheduled work and finished work appear in an Activity View.
You can also connect your own computer from the desktop app. It is off by default. With it on, a dot can work with local files and skills and fall back to your local browser when its cloud browser is blocked.
Rules and safeguards
Custom Rules set how a dot handles a type of action. There are four options:
| Option | What happens |
|---|---|
| Take action without asking | The dot does it |
| Take action if pre-approved | Only when you asked for that exact action in your prompt |
| Ask before taking action | The dot pauses for your approval |
| Hand off to you | The dot prepares the step and you finish it |
Some actions have fixed handling. Changing a password or transferring money hands control back to you. Permanently deleting data or installing software can require approval each time. Rules can tighten these defaults but not go below them.
Before planned actions run, a separate reviewer agent called auto-review checks them against your instructions, your Custom Rules and OpenAI's safety requirements, for example to catch an email going to the wrong recipient. On supported sign-ins, the dot pauses while you type credentials into a secure form, and they are passed to the browser without being shown to the model.
OpenAI published dot-specific test results in the GPT-6 Astra system card. In a bulk email test with 16,600 prompt-injection attacks among 50,000 messages, no attack succeeded. Scope violations, such as carrying information between unrelated tasks, rose from 8.6% to 19.7% when the number of intervening tasks doubled from five to ten. OpenAI notes that these scenarios were chosen to be difficult.
ChatGPT dots price and access
| Plan | Monthly price | Dot |
|---|---|---|
| ChatGPT Free, Go, Plus | $0 to $20.00 | Not included |
| ChatGPT Pro | from $100.00 | First dot included, not in the EEA, Switzerland or the UK |
| Business Premium | $125.00 per seat, or $100.00 on annual billing | First dot included in all supported regions |
| Enterprise, Edu, Healthcare | Per agreement | Beta, off until an admin enables it |
Dots are for users aged 18 and over, and the rollout is gradual, so an eligible account may wait several days.
Conversations with a dot do not count toward ChatGPT usage limits. At launch, OpenAI said heavier dot usage would not count toward eligible plan allowances for the first month and that per-plan usage terms would follow. Codex and ChatGPT Work tasks that a dot creates count as usual. OpenAI says you will later be able to add more dots and raise a dot's speed or monthly work allowance, without prices so far. Users on the OpenAI forum have reported a rest message after long voice calls with a dot on Pro.
Limits to know
- Each person gets one dot. Teams of dots are planned, and Enterprise pilots can run specialist dots.
- You cannot view or delete individual dot memories. Reset deletes the dot with its conversations, memories and schedules, while files, Codex threads and chats it created remain.
- At launch a dot has no standalone email address and cannot call you. Texting is a limited US Pro beta.
- OpenAI's Help Center says a dot can make mistakes, including when following your rules.
- OpenAI's dots are unrelated to Dot, an earlier companion app from the startup New Computer.
Frequently asked questions
What is ChatGPT dots?
Always-on agents inside ChatGPT. Each dot runs on GPT-6 Astra, has its own cloud computer and browser, connects to over 4,000 apps, keeps a memory and keeps working toward goals between conversations.
Are ChatGPT dots free?
The first dot costs nothing extra, but it comes only with ChatGPT Pro (from $100.00 a month) or Business Premium. Free, Go and Plus do not include dots.
Can I use ChatGPT dots in the UK or EU?
Not on ChatGPT Pro, which excludes the EEA, Switzerland and the UK. Business Premium workspaces get dots in all supported ChatGPT regions.
Where can I talk to my dot?
In ChatGPT on desktop, web and mobile, in Slack and Microsoft Teams, and by voice call inside ChatGPT. US Pro users may get a texting beta.
Do dots count against my ChatGPT limits?
Chatting with a dot does not. OpenAI waived plan allowances for heavier dot work during the first month after launch and said per-plan terms would follow. Codex tasks a dot creates count normally.
Can a dot spend money or change passwords?
It can buy with a card saved on a merchant's site after your approval. Changing a password or transferring money hands control to you to finish.
How do I delete what a dot remembers?
Individual memories cannot be edited. Reset deletes the dot together with its conversations, memories and scheduled tasks.
Sources
Read on Oct 22, 2026. Plan prices are USD per month and exclude taxes.