OpenAI introduced Dots at DevDay as personal agents meant to keep working on goals a user sets. Initial access is limited to specified paid ChatGPT plans and eligible markets. The launch gives OpenAI a persistent agent product, but its vision of several specialized Dots cooperating for one user is a future plan, not a feature available to everyone today.
OpenAI introduced Dots at its September 29 DevDay as personal agents designed to keep working on goals set by a user. The company presents them as a way to assign an ongoing task rather than send a single prompt and wait for one answer. The initial version lets a person name and customize one agent; several specialized Dots working together remain a later plan.[1], [2]
Dots uses OpenAI’s GPT-6 Astra model and can be launched from ChatGPT or Codex. OpenAI’s examples include a developer asking an agent to watch customer feedback and prepare fixes, and a scientist asking one to rerun an analysis when new experimental data arrives. These are proposed uses, not measured success rates for unattended work. A person can also message the agent through workplace tools including Slack and Teams.[1]
At launch, the described access is limited to certain Pro and Business Premium subscribers in eligible markets. Text messaging support is still forthcoming, and the broader vision of multiple cooperating agents is separate from this first offer. OpenAI is also working on integration with Microsoft’s Agent 365 security controls. The announced product therefore has a defined entry point and communication channels, while the reported account does not quantify its error rate in deployed tasks.[1]
