ReAct (Reasoning + Acting)
Definition
ReAct is a paradigm where the model alternates reasoning (what to do next, why) and acting (tool calls). The observation from the environment feeds back into the next reasoning step, forming a loop until the task is done.
It is the standard pattern for agents that use tools: each action is preceded by a thought, which reduces blind or repetitive tool use. Often combined with chain-of-thought (reasoning inside the thought) and with RDD when specs guide decisions.
How it works
Prompt format is Thought → Action → Observation → Thought → … → Final Answer. The user gives a task; the agent produces a thought (reasoning about what to do), then an action (e.g. tool call). The environment/tools return an observation, which is appended to the context for the next thought. The loop continues until the agent outputs a final answer. The model decides when to call tools and when to conclude, which reduces arbitrary or repetitive actions. The sequence diagram below summarizes this flow; frameworks like LangChain implement ReAct-style agents with tool registration and message handling.
Use cases
ReAct fits agent workflows where each tool call should be preceded by a clear reasoning step.
- Agents that use tools (search, calculator, API) with explicit reasoning
- Reducing arbitrary or repetitive tool calls by interleaving thought
- Debuggable agent behavior via visible thought–action–observation traces
External documentation
- ReAct: Synergizing Reasoning and Acting in LLMs (Yao et al.) — Original ReAct paper
- LangChain – ReAct agent — ReAct-style agents in LangChain