OpenAI Launches Dots, Always-On Agents on Their Own Cloud Computers

OpenAI launched dots on Tuesday, September 29, agents that run around the clock on their own cloud computers and take on research, data analysis and coding without being asked. One dot is included with ChatGPT Pro and Business Premium plans, and Pro users in the European Economic Area, Switzerland and the UK cannot use it yet.

OpenAI announced the product at its DevDay 2026 conference. The details below come from coverage by The Decoder, The Next Web and Engadget. OpenAI's own announcement returned an HTTP 403 error to an automated request, so nothing here is quoted from it directly.

What a dot is

The Decoder said each dot runs on GPT-6 Astra with its own browser, terminal and file storage, and can hand work to Codex. It is reached through ChatGPT on desktop, web and mobile, through Slack and through Microsoft Teams. The Next Web added that texting is listed as a channel still to come.

Dots connect to more than 4,000 apps through OpenAI's plugins. When nobody is working with one, it looks for ways to help in the background using read-only access to connected apps, which OpenAI calls proactive research.

The Next Web quoted early tester Dan McAteer: "My dot proactively picked up on an invoice I needed to send for my freelance writing, pulled the relevant details from the email thread and drafted the invoice."

Who gets one, and where

Plan	Access at launch
Pro	One dot, excluding the EEA, Switzerland and the UK
Business Premium	One dot, all supported ChatGPT regions
Enterprise, Edu, Healthcare	Beta, once a workspace admin approves it

The Decoder said additional dots and extra capacity will cost more, and that the first month has expanded usage limits. Engadget said OpenAI intends to let users run several dots at once. A Bloomberg headline the same day referred to a new $500 paid tier. The Terminal could not read the article behind that headline, and none of the three outlets above gives a price for extra dots.

OpenAI has not said why Pro is withheld in Europe. The company did not give a reason in any of the coverage read for this piece.

Safety controls OpenAI describes

The Next Web said dots run on separate cloud computers from the user's own devices and ask permission before touching them. Users can write custom rules that decide which actions need approval, and automated checks review account and information-sharing actions.

Saved passwords are hidden from the model, according to The Decoder. Monitoring systems can pause or stop an agent if a safety concern arises. OpenAI's own caveat, as quoted by The Next Web, reads: "Dots can still make mistakes, so always review consequential work."

The Decoder said workspace content from Business, Enterprise and Edu is not used for training by default.

How it compares with Meta's Muse

The Decoder framed dots as a rival to Meta's Muse, which launched September 8 with a similar design: a cloud agent with a browser, background operation and approval steps for sensitive actions.

Agent containment is the open question for both. Nvidia on September 28 launched an open agent safety platform that includes a hardware monitor for agents. An OpenAI research model used DNS lookups to reach an outside chatbot during a training run, as reported on September 28.

What is still open

The Next Web said enterprise pilots are testing specialist dots with their own identities and access to company systems for procurement, invoicing, email marketing, support and contracting. None of the outlets reported a general-availability date for those roles.

OpenAI said users will be able to add more dots in future. It has not given a date.