Updated on September 30, 2026
Dots are OpenAI's always-on personal AI agents, launched at DevDay on September 29, 2026, inside ChatGPT. Each Dot runs on its own cloud computer, is powered by GPT-6 Astra and keeps chasing your goals after the tab closes. The first one comes at no extra charge with Pro and Business Premium plans in eligible markets, though the Pro rollout skips the EEA, Switzerland and the UK. Meta's Muse got a head start of a few weeks, and OpenAI plainly wants the round back.
A Dot takes ownership of a goal and keeps working on it between your conversations, where agent mode in ChatGPT runs one task and stops. It is the textbook case of a program acting autonomously toward a goal, stretched into round-the-clock operation. OpenAI's launch examples are down to earth. A developer's Dot watches customer feedback and prepares bug fixes, while a scientist's Dot reruns analyses whenever fresh data lands.
Each Dot shows up as a small animated character you name and customize. OpenAI's own showcase features a yellow Dot named Alfred, and one early tester's Dot even spotted a forgotten invoice, prepared it and sent it.
The timing tells its own story. Meta's Muse arrived a few weeks earlier, so autonomous AI agents have become the main front in the race between the big labs.
Every Dot gets a dedicated cloud computer with its own browser and reaches over 4,000 apps through OpenAI's plugin catalog. You can open that virtual machine whenever you like and watch the agent at work (oddly soothing during the first days, truth be told).
The conversation follows you around. You message or call your Dot in ChatGPT on web, mobile and desktop, then pick up the same thread in Slack or Microsoft Teams, with texting announced as coming soon.
| Aspect | ChatGPT agent mode | A Dot |
|---|---|---|
| Duration | One-off task, launched on demand | Ongoing goal, worked 24/7 |
| Machine | Temporary session | Dedicated, inspectable cloud computer |
| Channels | The chat window | ChatGPT, Slack, Teams, texting soon |
| Memory | Session context | Learns from your feedback over time |
| Quotas | Counts against your limits | Dot chats excluded from ChatGPT limits |
Setup happens in the ChatGPT desktop app or a desktop browser, on a Pro or Business Premium plan in an eligible market, while Enterprise, Edu and Healthcare workspaces get a beta that stays off until an admin switches it on. Everyone starts with a single agent, and OpenAI plans multiple Dots per person, then teams of them cooperating over time.
Trust is the other half of the pitch. Background research runs read-only, an auto-review step screens consequential actions, and moves such as changing a password stay in human hands. An integration with Microsoft Agent 365 security controls is in development, and tasks handed to Codex or ChatGPT Work still count against those quotas.
Dots are always-on personal AI agents built into ChatGPT and powered by GPT-6 Astra. Each one appears as a small animated avatar, runs on its own cloud computer with a browser, and keeps making progress on your goals after you close the chat, messaging you when a decision is yours to make.
Agent mode runs a single task launched from the chat and then stops. A Dot holds onto a goal, works in the background on its own machine, learns from your feedback over time, and can carry its own identity and credentials. One follows an order, the other owns a lasting responsibility.
You create your first Dot from the ChatGPT desktop app or a desktop browser, on a Pro or Business Premium plan in an eligible market. Give it a name, connect your apps, then talk to it in ChatGPT, Slack or Teams. Enterprise, Edu and Healthcare workspaces get a beta that admins must enable.
Several guardrails are active by default. Background research is read-only, an auto-review step checks consequential actions, sensitive moves such as password changes stay with you, and a monitoring system can pause the agent. Caution is still warranted, since OpenAI itself held back GPT-6.1 Astra over safety concerns.
Verdict: Handing over a whole project instead of a single prompt is the habit Dots aim to build. Pro and Business Premium subscribers who already spend their day in ChatGPT, Slack and Teams will feel the difference within a week.
Manus AI Agent 
OpenClaw 
Hermes Agent 
