OpenAI's always-on agents can connect to multiple local machines, with users deciding which one handles each job
OpenAI’s always-on ChatGPT agents, called dots, now let users choose which of their local computers handles a given task.
What dots are and what changed #
OpenAI introduced dots on September 29, 2026, at its annual DevDay event. Each dot is an always-on AI agent that runs on a dedicated cloud computer powered by the GPT-6 series.
The launch models are GPT-6 Astra, with later models such as GPT-6.1 Terra also in the mix. The core idea is persistence: a dot can keep working on tasks, take in user feedback, and push projects forward between conversations.
The newest wrinkle concerns local machines. By default, dots live in an isolated cloud environment and have no access to a user’s own computer.
Users can switch that on through the ChatGPT desktop app. Once enabled, they can connect multiple local computers and designate which one runs a specific task.
Who gets access, and what it costs #
Dots are not available to everyone. Access is limited to ChatGPT Pro subscribers on the $100+ tiers, Business Premium customers, and Enterprise workspaces, with initial enterprise availability in select markets.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Users on the Free, Go, and Plus plans do not get dots.
For eligible Pro and Business Premium users, the first dot comes at no extra cost. Usage accounting has a split personality. Chatting with a dot does not count toward ChatGPT usage limits, but the work a user delegates to it does.
The plumbing behind the agent #
Alongside dots, OpenAI has connected ChatGPT to more than 4,000 applications through plugins. That gives each dot a large toolbox for moving work across the software people already use.
Dots can be reached through the ChatGPT apps and Slack. Microsoft Teams and text messaging are in limited testing.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.