ReAct (Reason + Act)
Definition
An agent architecture pattern where the model alternates between reasoning (thinking about what to do) and acting (calling tools or taking actions). In each iteration, the agent: (1) observes the current state, (2) reasons about what to do next, (3) selects and executes an action, (4) observes the result, and repeats. ReAct is the most common agent loop pattern and the foundation of most production agents. It outperforms pure chain-of-thought (reasoning only) and pure action (acting without reasoning) approaches.
Builder Context
ReAct is your default agent architecture. Implement it as a simple while loop: prompt the model with the task + history + available tools, parse the response for tool calls, execute them, append results to history, repeat until the model signals completion. The main tuning levers: max iterations (prevent infinite loops), tool selection prompt quality, and observation formatting. When ReAct agents fail, the issue is usually in the observation step — the agent gets poor or missing feedback from tool calls.