Always-on AI agent

A field term for assistants that keep working outside a single chat turn. Vendors use their own words. OpenAI describes Dots as always-on agents.

Also known as: always-on assistant · persistent agent · long-running agent · universal AI assistant

Official

OpenAI says dots are remarkably capable, always-on agents built to handle everything.
Source: OpenAI: Introducing dots
OpenAI says that when you are not actively working with it, the dot looks for ways to help using already-connected apps with read-only tools. In that proactive-research mode, OpenAI says it cannot send messages, change app content, or control the browser or computer.
Source: OpenAI: Introducing dots
Anthropic contrasts regular chat with Cowork: in regular chat, Claude responds to messages but cannot access files directly; in Cowork, Claude can read, edit, and create files in specified folders and complete tasks rather than only describe them.
Source: Anthropic: Claude Cowork FAQ
Anthropic says a Cowork task started from a phone keeps working in the cloud even when the laptop is closed, and the product page states “Close your laptop, it keeps going.”
Source: Anthropic: Claude Cowork
Google describes a vision of Gemini as a universal AI assistant and says Project Astra capabilities such as video understanding, screen sharing, and memory have been integrating into Gemini Live.
Source: Google: Gemini as a universal AI assistant
Google DeepMind describes Project Astra as a research prototype being refined by a limited number of trusted testers, and says it is working to bring Astra capabilities to Gemini Live and other surfaces.
Source: Google DeepMind: Project Astra
OpenAI’s announcement states: “Workspace agents are an evolution of GPTs.” OpenAI’s announcement also states: “Today, teams can interact with agents in ChatGPT and Slack, with more surfaces coming soon.”
Source: OpenAI: Introducing workspace agents in ChatGPT
Meta says that for tasks that take more time, Muse keeps working after people close the app, and comes back when something changes or when it needs approval, such as before sending an email or making a purchase.
Source: Meta Newsroom: Introducing Muse
Meta says Muse runs on Muse Secure VM, a dedicated virtual machine that houses both the agent and a person’s data, because “personal agents need a new kind of secure computer.”
Source: Meta Newsroom: Introducing Muse