OpenAI announces “Dots,” always-on AI agents that proactively carry out tasks for users by using a computer and connected applications. The company describes Dots as working toward user-set goals in the background—24/7—while learning preferences from user feedback over time. Sources also report OpenAI links the rollout to its recent GPT-6 Astra announcement.

The outlets agree that Dots launch with limited initial access. At the start, each user receives a primary Dot, with plans to expand to multiple agents over time. Dots run on a dedicated cloud computer with browser capability, and can connect to thousands of apps via OpenAI’s plugin ecosystem, enabling tasks such as drafting documents, conducting research, updating materials, and supporting software development. Users can launch a Dot from ChatGPT or coding tools, set permissions and rules for when the agent can act on its own, and require approval where appropriate.

Coverage differs slightly on timing and availability. One outlet says OpenAI rolls out Dots to paid users starting Tuesday, while another describes same-day availability tied to OpenAI’s DevDay announcement. Both note rollout is gradual and targeted at specific paid tiers, with an enterprise beta planned. Microsoft is referenced as a partner for bringing specialist Dots into an “Agent 365” offering.