Explainer
What Are OpenAI Dots? The Always-On Agents Explained
Dots are OpenAI's always-on agents: named helpers with their own cloud computer, browser, and standing goals. Here is what they are, what they can do, and what is still unknown.
Last verified: September 30, 2026 · Reviewed by AstraDot editorial desk
OpenAI Dots are always-on AI agents announced by OpenAI on 29 September 2026. You start with a single primary Dot, give it a name, and hand it ongoing work. Each Dot runs on its own cloud computer with its own browser, connects to 4,000+ apps, and is powered by the GPT-6 Astra model. Unlike a chat assistant, a Dot is built to keep going after you close the window.
What a Dot actually is
Think of a Dot as a named coworker with a standing brief. Instead of you opening a chat and describing one task, a Dot holds a goal, works toward it on its own, and reports back. OpenAI describes Dots as agents that can take on a project and run with it while handling several others, because each one has its own environment and its own memory of the feedback you give it.
The defining traits are simple to state: a Dot is named, always-on, isolated on a cloud computer, and connected to the tools it needs. Everything else — how fast it runs, how much it can do per month, and how many you can have — is where OpenAI has drawn the line on published detail.
A Dot vs a chat assistant
Both live inside ChatGPT, so they can feel similar at first. The differences show up in how work gets done.
| Aspect | A chat assistant | A Dot |
|---|---|---|
| Unit of work | One prompt, one reply | An ongoing goal you hand over |
| Runtime | Runs while you are in the chat | Works 24/7, including in the background |
| Environment | Shares your chat session | Its own cloud computer and browser |
| Concurrency | Answer to answer | Can run a project while handling several others |
| Tools | Whatever you attach per conversation | Included apps via OpenAI's plugin ecosystem; can write and test code |
| Learning | Mostly within a conversation | Learns from your feedback over time |
| Oversight | You read the reply | Auto-review, Custom Rules, and an activity view |
What a Dot can do
- Work on its own cloud computer. Each Dot has a machine separate from your own devices unless you deliberately connect them.
- Browse the web. A Dot has its own browser and can use it as part of a task.
- Write and test code. It can produce code and check that it runs, which is why the developer example below works.
- Connect to 4,000+ apps through OpenAI's plugin ecosystem, so it can act in the tools your team already uses rather than only talking about them.
- Learn from your feedback. A Dot is meant to improve as you correct it, rather than resetting every time.
- Run 24/7. It keeps making progress whether or not you are watching.
The examples OpenAI gave
At launch, OpenAI illustrated Dots with five concrete roles. They are useful because they show the shape of a real assignment rather than an abstract feature list.
| Role | What the Dot does |
|---|---|
| Developer | Turns customer feedback into tested pull requests. |
| Launch lead | Revises launch materials as the scope changes. |
| Scientist | Reruns analysis and updates the figures. |
| Sales lead | Updates a proposal and builds a proof of concept. |
| Content creator | Turns an interview transcript into clips, show notes, and posts. |
Where you reach your Dot
Channels matter, because a Dot is meant to be reachable where you already work:
- ChatGPT — desktop, web, and mobile.
- Slack and Teams, where it can work alongside a team.
- Texting over iMessage/RCS, coming on a limited Pro waitlist.
A Dot can also message you first. OpenAI says Dots carry context across these channels, so a conversation you start on desktop can continue from your phone.
How you stay in control
Always-on agents only work if people trust them, so OpenAI paired the capability with several controls:
- Isolation. A Dot is on its own cloud computer, separate from yours unless you connect the two.
- Passwords stay hidden. Saved passwords are used without being exposed to the model.
- Read-only proactive research. When a Dot researches in the background, it uses read-only tools: it cannot send messages, change app content, or control your browser or computer.
- Auto-review. Actions are checked against your instructions, Custom Rules, and safety requirements before they go ahead.
- Sensitive tasks stay with you. Some actions, such as changing a password, are always reserved for the human.
- Activity view. You can see progress, including background work.
- OpenAI can intervene. If monitoring finds a concern, OpenAI can pause or stop a Dot.
Dots can make mistakes. Treat consequential or irreversible work as something to review before you rely on it.
Who gets a Dot
Dots are rolling out to Pro and Business Premium in eligible markets, with the first Dot included at no extra cost. Enterprise plans, including Edu and Healthcare, can try the beta when a workspace admin enables it. OpenAI says Dots will expand to more users "soon" and has not published a date. See how to get a Dot for the eligibility table.
What is still not published
This guide will not guess. As of OpenAI's launch announcement, several figures are simply not public:
- The price of additional Dots beyond the first.
- How speed and monthly work can be scaled, and what that costs.
- The exact size of the "deeper work" allowance included with a plan.
- A date for the wider rollout beyond Pro, Business Premium, and opted-in beta workspaces.
- Benchmark performance for GPT-6 Astra (see the model page).
Source: OpenAI's Introducing Dots announcement and Dots help center. This site is independent and not affiliated with OpenAI.