A personal AI agent sounds like software living on your computer. OpenAI’s Dots take the opposite route: each agent runs on its own cloud PC, shifting the hardware burden away from the user while adding a familiar dependency on internet access and OpenAI’s services.
An agent with its own computer
OpenAI describes Dots as always-on agents that learn what matters to each user and take on work on their behalf. Each one is powered by GPT-6 Astra and runs on a cloud virtual PC hosted on OpenAI’s servers, rather than on the user’s device.
That setup means users do not need a powerful local machine to run an agent. It also means the agent depends on a working connection and OpenAI’s infrastructure. The computer is elsewhere, though users can log in to its virtual PC and inspect the agent’s work.
OpenAI presents Dots as a new way to work with AI, but the report does not describe specific tasks the agents can complete. For now, the clearer details are about how they run and how users can reach them.
Connections and access
Software connections: OpenAI says a plugin ecosystem lets Dots connect to more than 4,000 programs.
Ways to interact: Users can communicate with an agent through ChatGPT, SMS, a messaging app, or voice calls.
Plans and rollout: OpenAI is beginning access for Pro, Business Premium, and Enterprise customers in supported countries and regions. It is also offering a limited preview to enterprise customers for specialized Dots with their own accounts and defined responsibilities within an organization.
Users can start with one Dot, give it a name, and configure it. OpenAI says it eventually expects people to use teams of Dots working together. For now, the practical trade-off is straightforward: the agent gets a computer of its own, and the user gets another service that needs to stay online.