Agent Memory and Context

A Long Task in a Finite Window

An agent working a real task builds up history fast.

Think of a two-hour debugging session, or a support case with fifty tool calls behind it.

The context window is the model's working memory: the text it can read in a single call.

This history may exceed the useful context budget, even before it reaches the model's maximum window.

Repeated history adds input tokens and latency, as described in the context-cost lesson