OpenAI began rolling out dots in ChatGPT on September 29, 2026, for Pro and Business Premium users in supported markets. The company describes the AI agents as ongoing project collaborators that can handle several assignments at once, without requiring users to manage a separate conversation for every task.
A cloud computer and context across channels
According to OpenAI, dots run on GPT‑6 Astra and have their own cloud computer, allowing them to keep working toward a user’s goals. The company says its plugin ecosystem connects the agents to more than 4,000 applications. Those connections supply tools for working in the services a user connects to the agent.
OpenAI says users can message or call their dot through ChatGPT on desktop, web and mobile, and exchange messages with it in Slack and Teams. The agent is designed to carry project context between those channels, including when a project starts in ChatGPT and continues with a team in Slack. SMS communication is planned.
Communication can also begin with the agent: OpenAI says a dot can send progress updates, ask questions or bring decisions back to the user. Through continued collaboration, the company says it learns the user’s preferences and expectations for a finished result.
Workflows for code, research and content
For developers, OpenAI describes a workflow that starts with customer feedback rather than an individually assigned coding task. A dot watches for recurring requests, defines smaller improvements and bug fixes, then implements and tests them. It presents completed pull requests for human review, with accompanying videos showing the changes.
For scientific work, the company describes an agent that reruns analyses when new data arrives, investigates unexpected results and updates figures for a paper. If the findings change, the dot is intended to revise the accompanying explanation as well. It flags the portions that require the researcher’s review.
OpenAI’s content-creation example begins with a new interview transcript. The dot selects passages suitable for short clips, prepares show notes and drafts social posts for approval. User edits are then meant to carry across the related materials, rather than remaining confined to one draft.
Inspecting work and setting action permissions
Users can open their agent’s cloud computer at any time to inspect its work. A dot can also connect to other devices, but using the user’s own laptop requires permission.
OpenAI calls the agent’s background search for useful work “proactive research.” In that mode, it uses read-only tools in applications the user has already connected. It cannot send messages, modify application content or control the browser or computer while performing that research.
Custom Rules let users allow particular actions, require approval or block them. An auto-review check assesses actions that could affect accounts or share information against the user’s instructions and safety requirements. Some sensitive tasks, including changing a password, must always be performed by the user.
How workspace and personal data are treated
OpenAI says it does not use content from Business, Enterprise or Edu workspaces to improve its models by default. For personal plans, user settings determine whether conversations and work with dots can be used for that purpose. The company does not train directly on proactive research or the agent’s internal notes, although information from them may enter an eligible conversation or task, depending on those settings.
Specialist dots start with enterprise pilots
Alongside the personal agent, OpenAI introduced specialist dots for defined business tasks through focused enterprise pilots. Each organization assigns its agents their own identities, credentials and access to the systems they need. OpenAI’s teams will work with participating organizations to define responsibilities, available tools and the arrangements for human review and approval.
OpenAI is also working with Microsoft on an integration between specialist dots and Agent 365’s enterprise governance and security controls. The planned integration is intended to let businesses manage these agents through Microsoft tools they already use.
Access, setup and usage limits
Enterprise users, including Edu and Healthcare, can try the beta when their workspace administrator enables it. The first dot is included in Pro and Business Premium at no additional charge. Those plans also include an allowance for more intensive work, with expanded limits during the first month after launch.
Initial setup requires the ChatGPT desktop application or a desktop browser, where users create their first dot and connect applications. After setup, they can message it from the ChatGPT mobile app. Conversations with a dot do not count toward ChatGPT usage limits, but tasks it starts or manages in Codex or ChatGPT Work count against those services’ applicable limits as usual.
