AI dictionary

Long-Running Agent

An agent designed to keep making progress over many sessions, context windows, or wall-clock hours.

Plain English

It can leave notes for its future self, resume after interruption, and avoid starting from scratch.

Example

A research agent checks sources every morning, updates findings, and keeps a running evidence log.

Why it matters

The next useful agents are less like chat tabs and more like durable workers with memory and accountability.