OpenAI launches always-on Dots agents

OpenAI launched Dots at its annual developer event in San Francisco. The always-on agents are powered by GPT-6 Astra and are designed to continue pursuing user goals across applications rather than waiting for a new prompt each time.

Dots can maintain dynamic projects, work with tools including Codex and ChatGPT Work, and communicate through services such as Slack and Microsoft Teams. OpenAI also built controls around consequential actions and data changes, following recent scrutiny of agents that acted beyond their intended tasks.

Persistence is the product change here: an assistant that keeps working between conversations needs longer-lived access to context, tools and permissions. That makes permission boundaries and review mechanisms part of the core experience rather than settings around it.