OpenAI introduced Dots, a new system of persistent AI agents integrated into ChatGPT, during the DevDay conference. These agents are powered by the GPT-6 Astra model and operate on dedicated cloud computers. Users can monitor the activity of these agents in real time via their individual cloud environments.
The agents are designed to execute complex tasks autonomously across multiple projects. They can integrate with over 4,000 applications via plugins and access user laptops with explicit permission. Interaction is supported across desktop, web, mobile, Slack, and Microsoft Teams, with context maintained seamlessly between these platforms.
Safety features include an auto-review system for actions affecting user accounts and a monitoring mechanism capable of pausing agents. During proactive research, agent access to user applications is restricted to read-only mode. OpenAI advises users to verify all output, noting that the agents may still produce errors.
OpenAI also previewed specialist agents for corporate roles like procurement and invoice processing. These are managed through Microsoft Agent 365 governance controls. Customers can initiate focused pilot programs where OpenAI engineers assist in defining specific agent responsibilities and workflows.
Regarding pricing, one agent is included at no additional cost with Pro or Business Premium plans. Usage will not count against plan limits during the first month, with future pricing for additional agents or increased capacity yet to be announced. This launch positions OpenAI against Meta’s Muse agent.
