Pattern
The Agent Loop
Observe → Think → Act → Repeat. The simplest possible agent is this loop.
When to use
When you need an agent at all. The loop is the abstraction; everything else (tools, memory, multi-agent topologies) is layered on top.
When not to use
When a single LLM call answers the question. The loop adds latency and cost; only pay it when the task genuinely requires iteration.