OpenAI introduces dots, personal agents that work around the clock inside ChatGPT

OpenAI introduced personal agents called dots at its DevDay event on September 29, 2026. Dots are designed as helpers that live inside ChatGPT, get work done on the user's behalf and can run at any hour. According to the company, dots run on GPT-6 Astra and learn the user's habits over time through feedback.

What a dot does

Each dot has its own cloud computer and browser. It can connect to thousands of apps through OpenAI's plugin system; users can open the dot's computer to watch what it is doing step by step, and with permission a dot can also connect to the user's own laptop. Dots can be reached from the ChatGPT desktop, web and mobile apps, and through Slack and Microsoft Teams by message or voice call; SMS support is coming soon. Users give their first dot a name, and teams of several dots will be possible later.

When the user is away, a dot can research only with read-only tools; it cannot send messages or change content. It can use saved passwords without revealing them to the model. Users can set custom rules for which actions are allowed, need approval or are blocked, and everything shows up in an activity view. Risky actions go through automatic review, and sensitive tasks such as changing a password always stay with the user. According to OpenAI, a monitoring system can stop the work if it spots a problem.

Specialist dots for companies

OpenAI also announced specialist dots, which take on a specific role inside an organization with their own identity and permissions. They start as focused enterprise pilots for tasks such as procurement, invoice processing, email marketing and customer support. An integration with Microsoft Agent 365 is planned, so these dots can be managed with Microsoft's governance and security tools. According to the company, Business, Enterprise and Edu content is not used to train models by default.

Who can use it

Dots are rolling out gradually to Pro and Business Premium users in "eligible markets"; the list of markets was not specified. In Enterprise, Edu and Healthcare plans they are available as a beta if an admin turns them on. Setup is done from the desktop app or a browser, after which a dot can also be managed from mobile. Current plan and pricing details are on OpenAI's site.

Any agent that connects to apps and acts for the user can cause real consequences with a wrong step. At first it is wise to give a dot narrow permissions, keep approval rules strict and read the activity log regularly. Companies should review their own data policies before connecting apps that hold personal data or trade secrets. To see how agents behave in real use, it is worth following evaluations from independent security researchers.

Sources: OpenAI: Introducing dots, OpenAI: DevDay 2026 recap, ChatGPT: dots product page.