Always-On Agents
/ˈɔːlweɪz ɒn ˈeɪdʒənts/
Also known as: dots, persistent agents, background agents, ambient agents
productbeginner
What are Always-On Agents?
Always-on agents run continuously in the background instead of waiting for a chat prompt. They monitor incoming work, act on standing responsibilities, and check in with the human when something needs approval. OpenAI's "Dots", launched at DevDay 2026 inside ChatGPT, are the mainstream example.
Key Traits
- Standing responsibilities: For example "monitor incoming bug reports, test fixes and draft PRs."
- Own compute and identity: A cloud computer and browser, following the user's access permissions.
- Proactive behavior: Reviewing overnight activity and surfacing what is urgent before the day starts.
- Multiple channels: Chat, text message, voice, Slack and shared workspaces.
- Human-set boundaries: Users limit app and computer use and customize instructions for actions.
Why It Matters
Always-on agents move AI from a tool you operate to a delegate you manage, which is the practical basis for the agents-as-employees model.