OpenAI launches Dots, always-on agents for ChatGPT Pro and business plans

By EnkiEdited by VK, Editor

Published 30 Sep 2026, 11:32 UTC

Reporting from TechCrunch, OpenAI, The Decoder, WIRED, The Verge

OpenAI unveiled Dots at DevDay: persistent GPT-6 Astra agents with their own cloud computer that work in the background across connected apps. One is included on Pro and Business Premium, with extra capacity to be sold later.

What it means for founders

Budget for usage, not seats. The chat is included, but anything a dot builds in Codex draws on the same limits your team already spends, so a busy dot can quietly exhaust them.
Treat app access as a security decision. A dot reading your inbox, CRM and repo all day is a new standing credential. Start it on narrow connections, set Custom Rules before it acts, and prefer a Business tier if training on your data is a concern.
Agent features in your product now compete with a bundle. If you sell a background assistant for triage, invoicing or support, your buyers now get one inside a plan they may already pay for. Depth, vertical data and audit trails are the defensible ground.
Watch for OpenAI's price for extra dots and the end of the launch month limits, which will show what always-on work really costs.

The story

OpenAI used its DevDay keynote on Tuesday to launch Dots, persistent agents that keep working on a user's goals in the background instead of waiting for a prompt. Each dot runs on GPT-6 Astra with its own cloud computer and browser, and OpenAI says plugins let it reach more than 4,000 apps. OpenAI's announcement limits each user to one dot for now.

How it works

A dot is reachable in ChatGPT on web, desktop and mobile, by chat or voice, and inside Slack and Microsoft Teams, carrying the same context across all of them. Texting is promised but not live. Users can open the dot's cloud machine to watch or take over, and can separately allow it onto their own laptop. In one OpenAI example, a tester's dot noticed an unbilled client, drafted a bill and mailed it once he signed off.

The launch lands three weeks after Meta's Muse, which works in a similar way and has drawn early complaints about acting without its owner's knowledge.

Plans, pricing and access

Dots are rolling out to ChatGPT Pro and Business Premium subscribers in what OpenAI calls eligible markets, and Enterprise, Edu and Healthcare workspaces can try a beta once an admin turns it on. WIRED puts Pro at $100 a month. One dot is included, chatting with it does not use up ChatGPT limits, and work it hands to Codex or ChatGPT Work is metered as usual. OpenAI says extra dots, more speed and more monthly work volume will be sold later. The Decoder reports that Pro subscribers in the EEA, the UK and Switzerland are excluded for now, while Business Premium covers all supported regions.

Data and permissions

Users pick which apps a dot can touch through ChatGPT's existing app controls. When idle, it browses those apps with read-only tools, which OpenAI calls proactive research. Built-in rules, user-set Custom Rules and an automated review decide which actions go ahead, which need sign-off and which are blocked; password changes always stay with the human, and saved logins are used without the model seeing them. OpenAI says it does not train on Business, Enterprise or Edu content by default, and personal plan users choose whether their dot's work is used.

For companies, OpenAI is piloting role-specific dots that carry separate credentials, and is working with Microsoft to manage them through Agent 365.

What we don't know yet

OpenAI has not priced additional dots, published the size of the included work allowance or said when the boosted first month ends. The texting date is open, and there is no public data yet on how often dots get tasks wrong.