OpenAI introduced dots on September 29, describing them as always-on AI agents that can continue working on a user's goals across tasks. A dot has its own cloud computer and can use connected apps, a browser and, with permission, other devices. The initial rollout is for eligible ChatGPT Pro and Business Premium users, with an Enterprise beta controlled by workspace administrators.
The launch marks a shift from asking a chatbot for a single answer toward assigning continuing work to an agent. That does not mean every example in the announcement is immediately available to every account. Access depends on plan, eligible market, connected apps and the permissions granted by the user or organization.
What is a dot?
A dot is an agent that can take on projects, keep context and make progress between conversations. OpenAI says it is powered by GPT-6 Astra and works on a separate cloud computer. A person can inspect that computer's activity and give feedback as the work proceeds. OpenAI presents tasks such as investigating a reported software bug, revising launch materials and updating analysis when new evidence arrives as examples of possible workflows.
The first experience is a primary dot that a user can name. OpenAI is separately previewing specialist dots for organizations, with their own identities and narrowly assigned responsibilities. Those enterprise pilots are distinct from the consumer and business rollout of the primary dot.
OpenAI's launch announcement explains the product and its limits. Its dots product page provides another starting point for availability and setup details as the rollout changes.
Who can use dots and how do they start?
OpenAI says dots are beginning to roll out to Pro and Business Premium users in eligible markets. Enterprise users, including Edu and Healthcare workspaces, can try the beta when an administrator enables it. The first dot is included with Pro or Business Premium at no extra charge, though deeper work has an allowance. An included agent therefore should not be read as unlimited task execution.
Initial creation takes place in the ChatGPT desktop app or a desktop browser. After setup, a user can message the dot from the mobile app. OpenAI also describes communication through Slack and Teams, depending on connected services. Availability and supported actions may evolve, so the current product screen is the final check for a particular account.
What can it access, and who approves actions?
An agent's useful range depends on access. Users choose which apps to connect and can manage app permissions in ChatGPT. A dot's cloud computer is separate from the user's own computer unless that device is explicitly connected. OpenAI says background proactive research uses already connected apps with read-only tools when the user is not actively working with the dot.
Actions that affect accounts or share information are subject to OpenAI's action review and approval rules. Users can set Custom Rules to permit, require approval for or block particular actions, within built-in safeguards. Activity View lets them follow progress and redirect work. Those controls matter more than a broad claim that the system works around the clock: always-on work still operates within permissions and review boundaries.
For Business and Enterprise workspaces, OpenAI says content is not used to improve models by default. Personal-plan settings differ and should be checked in account controls. The distinction is important for teams evaluating whether an agent should connect to work data.
What changes for everyday users?
The practical test is whether a dot can bring a finished, inspectable piece of work back to its owner, rather than merely produce a draft answer. OpenAI's examples include project updates, research and code changes, but actual performance will vary with access, task clarity and review. Start with a bounded task, check the resulting work and adjust permissions to match what the agent needs.
Is a dot available on the free plan? OpenAI's launch announcement names Pro and Business Premium for the initial rollout, plus an admin-enabled Enterprise beta. It says expansion to more users is planned, without promising a date.
