ReAct — the thought–action–observation loop

ReAct (Reason + Act) is the pattern that made tool-using agents practical and
remains the default agent loop in most frameworks. It interleaves reasoning traces
with tool actions and their observations in a repeating Thought → Action →
Observation loop, so each external result updates the agent's understanding before
it decides the next step. The key distinction taught here: CoT is sufficient for
pure reasoning, but ReAct is necessary the moment the task requires external data
or actions. ReAct is the simplest effective control flow when each next step
depends heavily on what the previous step returned.