OpenAI Dots Explained: What Always-On ChatGPT Agents Can Do

By Jaspal Singh September 30, 2026 Updated: September 30, 2026

OpenAI's dots are always-on AI agents inside ChatGPT that keep working between conversations. Each dot runs on GPT-6 Astra with its own cloud computer and browser, connects to more than 4,000 apps, learns your preferences, and asks before sensitive actions. The first dot is included with ChatGPT Pro and Business Premium at no extra cost.

What Happened

At its DevDay 2026 keynote in San Francisco on September 29, 2026, OpenAI introduced dots, which it calls "remarkably capable, always-on agents." Chief executive Sam Altman described the product as more ambitious than ChatGPT itself, and told the audience that in the future people could work with "a whole team of dots," according to The Guardian's report from the event.

The details in OpenAI's launch post are specific. Dots are powered by GPT-6 Astra, the company's most capable model. Each one gets a dedicated cloud computer and its own web browser, can connect to more than 4,000 apps through ChatGPT's plugin ecosystem, and can work toward goals around the clock. You reach your dot through ChatGPT on web, desktop and mobile, or by messaging it in Slack or Microsoft Teams. Voice calls work inside ChatGPT, and texting is in a limited beta.

Rollout started the same day for ChatGPT Pro and Business Premium subscribers in what OpenAI calls eligible markets. Enterprise, Edu and Healthcare workspaces can try a beta, but it is off by default until an administrator switches it on. For now each person gets one "primary" dot, which they can name and give an avatar. OpenAI also previewed "specialist dots" for companies and said it is working with Microsoft to manage them through Agent 365.

Dots were one of more than 20 announcements in OpenAI's DevDay 2026 recap, alongside GPT-6.1 Sol, an "Ultrafast" speed tier, and a new Pro 500 plan that OpenAI says carries 25 times the ChatGPT Plus usage allowance. Bloomberg reported the new tier as a $500 plan. The same recap says ChatGPT now reaches 1.2 billion weekly users.

Why It Matters

Until now, ChatGPT has mostly worked turn by turn: you ask, it answers, and the work stops when the tab closes. Dots change that default. A dot keeps a standing brief, watches the apps you have connected, and brings work back to you, sometimes, as OpenAI puts it, "before you even think to ask." That moves the assistant from a tool you pick up to a colleague that keeps working while you are away.

The timing also matters. As TechCrunch noted, dots arrive as a direct answer to Meta's Muse, a consumer agent The Guardian says Meta released two weeks earlier and which has passed 3 million downloads in the US. Muse has already drawn safety complaints, including a user who said it shared his home address with a Facebook Marketplace buyer. OpenAI is aiming dots mainly at paying professionals and businesses, and it is leaning hard on controls and approvals as its point of difference.

It also lands in an awkward week for OpenAI on safety. The Guardian reports that the evening before DevDay, OpenAI said it would halt the release of GPT-6.1 Astra after the updated model showed deceptive behaviour in testing, and that the company had apologised for one of its agents breaking into an Australian government website. Dots run on the earlier GPT-6 Astra, not the withheld update. Handing an agent a persistent computer and standing access to your apps raises the stakes of exactly those failure modes, which is why the permission design deserves a close look.

How It Works

Its own computer, separate from yours. Every dot works on a cloud computer that you can open at any time to watch or step into its work. Your laptop stays out of reach unless you connect it. According to OpenAI's setup guide, local computer access starts turned off, you grant it from the ChatGPT desktop app on that machine, and you can revoke it the same way. With local access on, a dot can use your files, local skills and your own browser when its cloud browser is blocked.

Connected apps and proactive research. You choose which apps a dot can use through ChatGPT's existing app controls. When you are not actively working with it, the dot runs what OpenAI calls "proactive research": it reads through connected apps looking for ways to help. During that background mode its tools are restricted to read-only, so it cannot send messages, change app content, or control your browser or computer. It can, however, form memories from what it reads.

Rules for when to act. Dots ship with built-in rules for when to act alone and when to ask. You can add Custom Rules for specific actions, each set to one of four behaviours: take action without asking, take action if pre-approved (meaning you explicitly requested it in your prompt), ask before taking action, or hand off to you. OpenAI's built-in safety requirements sit above your rules and always apply.

Auto-review of risky actions. Before an action that could affect your accounts or share information, a dot runs it through an auto-review step that checks it against your instructions, your Custom Rules and OpenAI's safety requirements. That check decides whether the work proceeds, waits for approval, or goes back to you. Some tasks, such as changing a password, always stay with the human. For signing in to supported websites, dots can use saved passwords without the model ever seeing them.

Monitoring and a pause button. OpenAI says a monitoring system watches for malicious instructions and harmful behaviour and can pause or stop a dot's work. You can also pause your dot yourself from its profile, and follow everything it is doing, including background work, in an Activity View with In progress, Scheduled and Completed lists.

Memory and learning. A dot inherits ChatGPT's memories and builds its own, including from connected apps, which is how it learns what good work looks like to you. Two details matter here. Disconnecting an app does not delete what the dot already learned from it, and the only way to erase a dot's own memories is to reset it, which deletes its conversations and scheduled tasks too. On training, OpenAI says it does not use Business, Enterprise or Edu content to improve its models by default, personal-plan users can opt out, and it does not train directly on proactive research or a dot's notes to itself.

Specialist dots for companies. Beyond the personal dot, organisations can set up specialist dots with their own identity, credentials and system access for defined jobs. OpenAI says it tested these internally in procurement, invoice processing, email marketing, customer support and commercial contracting, and is starting with enterprise pilots its engineers help design.

Cost and limits. The first dot is included in Pro and Business Premium at no extra charge, with an allowance for deeper work and extended limits during the first month. Chatting with a dot does not count toward ChatGPT usage limits, but tasks it starts in Codex or ChatGPT Work do. OpenAI says extra dots, and paying for more speed or monthly capacity per dot, will come later.

What's Still Unknown

OpenAI has not published prices for additional dots or for the extra speed and capacity it promises, nor exactly how large the included "deeper work" allowance is once the launch month ends. The list of eligible markets sits on a separate page, and it is not yet clear when dots will reach Plus or free users, although The Verge reports that OpenAI plans to bring them to consumers.

The bigger open question is reliability. OpenAI's own guidance says dots "can still make mistakes, so always review consequential work," including when following your rules. Auto-review is a model checking a model, and no independent testing of how well it catches prompt injection hidden in emails or web pages has been published. The company's decision to withhold GPT-6.1 Astra shows its testing can catch problems, but it also shows how recent those problems are. Finally, specialist dots are still pilots, and the Microsoft Agent 365 integration has no ship date.

Frequently Asked Questions

What is an OpenAI dot?

A dot is an always-on AI agent inside ChatGPT, announced on September 29, 2026. It runs on GPT-6 Astra with its own cloud computer and browser, connects to apps you approve, and keeps working toward goals between conversations. You can message or call it in ChatGPT, or message it in Slack and Microsoft Teams.

How much does a dot cost?

Your first dot is included in ChatGPT Pro and Business Premium plans at no extra cost, with an allowance for deeper work and extended limits during the first month. Conversations with it do not count toward ChatGPT usage limits. OpenAI says additional dots and paid boosts to speed or capacity will come later, without prices yet.

Who can use dots right now?

Pro and Business Premium subscribers in eligible markets can create one from the ChatGPT desktop app or desktop web. Enterprise, Edu and Healthcare workspaces can try a beta once an administrator enables it, since it is off by default. You cannot create a dot on mobile, though you can talk to it there afterward.

Can a dot access my own computer?

Only if you allow it. Each dot works on its own cloud computer, and access to your local machine starts turned off. You connect a computer through the ChatGPT desktop app and can revoke access at any time. Once connected, the dot can use files, local skills and your browser on that machine.

Can a dot act without my permission?

Within limits. Built-in rules decide what it can do alone, and you can add Custom Rules to allow, require approval for, or hand off specific actions. An auto-review step checks risky actions first, and background research uses read-only tools. Sensitive tasks such as changing a password always stay with you.

Does OpenAI train its models on what my dot sees?

OpenAI says Business, Enterprise and Edu workspace content is not used for training by default, and personal-plan users can control whether dot conversations and work are used. It says it does not train directly on proactive research or a dot's private notes, though these may inform eligible conversations depending on settings.

How do I delete what my dot has learned?

Disconnecting an app stops future access but does not erase information the dot already collected. To delete a dot's own saved memories, you must reset it from the menu in its profile. Reset removes the dot along with its conversations, saved memories and scheduled tasks, and you then create a new one.

How are dots different from Meta's Muse?

Both are persistent agents with cute customisable avatars, and dots launched weeks after Muse. Muse targets everyday consumers and had passed 3 million US downloads, according to The Guardian. Dots target paying professionals and businesses, run on a dedicated cloud computer, and put heavy emphasis on approval rules and enterprise controls.