OpenAI has introduced dots, a new kind of always-on AI agent designed to keep working after a chat ends. Announced on September 29, 2026, each dot runs on its own cloud computer, has a browser, and can use apps that its owner chooses to connect.
The product is powered by GPT-6 Astra and is beginning to roll out in ChatGPT for Pro and Business Premium subscribers in eligible markets. Enterprise customers, including Education and Healthcare workspaces, can try a beta when an administrator enables it. OpenAI says the first dot is included with Pro and Business Premium at no additional charge, although the plans include an allowance for more demanding work rather than unlimited compute.
What makes a dot different from a chatbot?
A normal chatbot waits for a prompt. A dot is intended to stay attached to ongoing projects, keep track of goals and watch for work that needs attention. OpenAI says a dot can handle several projects at once, learn from feedback and report back through ChatGPT, Slack or Microsoft Teams. Text-message support is planned but is not available at launch.
Each dot has a separate cloud computer and browser. Owners can inspect that computer while the dot is working, and they can optionally allow it to connect to a personal laptop. OpenAI says the agent can connect through its plugin ecosystem to more than 4,000 apps, but actual usefulness will depend heavily on which services are connected and what permissions are granted.
The company’s examples include watching customer feedback, preparing tested code fixes, updating research analysis when new data arrives, revising launch materials after a product change and preparing content from interview transcripts. These are examples supplied by OpenAI, not results from independent hands-on testing.
Rollout, price and usage limits
Dots started rolling out on September 29 to ChatGPT Pro and Business Premium users in eligible markets. The first dot is included at no extra charge. OpenAI says conversations with a dot do not count against ordinary ChatGPT message limits, but work that the dot starts or manages in Codex or ChatGPT Work still counts toward the limits of those products.
OpenAI has not published a standalone price for additional dots. It says future options will let people add more dots and pay for greater speed or a larger monthly work allowance. The company is also running focused enterprise pilots for “specialist dots,” which can have their own identities, credentials and access to internal systems.
Setup begins in the ChatGPT desktop app or a desktop web browser. After setup, owners can also message the dot from the ChatGPT mobile app. Availability is gradual, so an eligible subscription does not necessarily mean the feature will appear immediately.
OpenAI is putting guardrails around background work
An agent that can keep working and use connected accounts introduces more risk than a chat window. OpenAI says background “proactive research” is limited to read-only tools: in that mode, a dot cannot send messages, modify app content or control a browser or computer. A user chooses which apps the dot can access and can change those permissions through ChatGPT’s existing app controls.
Dots also use rules to decide when they can act independently and when they need approval. Custom Rules can allow, block or require approval for specified actions, while an Activity View shows background work. OpenAI’s auto-review system checks actions that could affect accounts or share information, and certain sensitive tasks—such as changing a password—must remain with the user.
Business, Enterprise and Education workspace content is not used to improve OpenAI’s models by default. Personal-plan users can control training through their data settings. OpenAI also cautions that dots can still make mistakes and says consequential work should be reviewed.
Why this launch matters
Dots move OpenAI beyond the familiar “ask a question, get an answer” model. The goal is a persistent software coworker that notices changes, continues multi-step assignments and asks for help only when needed. That could be useful for recurring work that currently falls between calendar reminders, project-management tools and human assistants.
The trade-off is that persistent agents need broad context and, in some cases, access to important accounts. The practical test will be whether dots can stay reliable over long-running projects without creating more review work than they save. Clear permissions, narrow initial tasks and careful approval rules will be sensible starting points for early users.
OpenAI’s official announcement has the full rollout and safety details. The company’s deployment safety update adds technical context, while Reuters reported on the launch from DevDay.