ReAct Pattern
definition
The ReAct (Reasoning + Acting) pattern is the foundational architecture where agents alternate between thinking steps and tool calls in a loop, combining chain-of-thought reasoning with grounded actions. Each cycle produces a thought (the model's reasoning about what to do), an action (a tool call or output), and an observation (the result of the action) that feeds the next iteration.
The ReAct (Reasoning + Acting) pattern is the foundational architecture where agents alternate between thinking steps and tool calls in a loop, combining chain-of-thought reasoning with grounded actions. Each cycle produces a thought (the model's reasoning about what to do), an action (a tool call or output), and an observation (the result of the action) that feeds the next iteration. This pattern is used by most modern agent frameworks because it provides transparency — you can inspect the reasoning chain to understand why the agent chose a particular action — while maintaining the flexibility to handle novel tasks the agent wasn't specifically programmed for. Understanding ReAct is critical because it is the default mental model for how agents work, and recognizing its limitations (unbounded loops, reasoning drift, error compounding) is what drives the need for more constrained patterns like state machines. This concept connects to the agent loop as the execution mechanism, chain of thought as the reasoning component, trace analysis for debugging ReAct chains, and state machine agents for the more constrained alternative.