Glossary
ReAct
An agent loop where the model alternates between reasoning steps (thought) and acting steps (tool call), explicitly interleaving free-form deliberation with structured tool use.
The pattern that proposed combining chain of thoughtagentsA prompting and training technique where the model emits step-by-step intermediate reasoning before its final answer, improving accuracy on multi-step problems. Open full entry with tool useagentsThe general pattern of an LLM invoking external functions, APIs, or systems to fetch data or take action, distinct from generating an answer purely from its weights. Open full entry in one explicit loop. The model emits a Thought (free-text reasoning), then an Action (a tool-call loopagentsThe host-orchestrated cycle behind tool use: the model emits a tool call, the host (not the model) executes it, and the result is fed back for the next turn. Open full entry ), gets an Observation (the tool result), and repeats. The thought step is what distinguishes ReAct from pure tool calling: it gives the model a place to plan, hypothesize, and self- correct between actions.
Most modern agent frameworks implement some descendant of ReAct. LangChainagentsThe earliest widely-adopted LLM agent and RAG orchestration framework (2022), now with the LangGraph extension for stateful multi-step agent workflows. Open full entry ’s classic agent, AutoGenagentsA Microsoft Research framework for multi-agent systems, with a conversation-pattern API for orchestrating multiple specialized agents to solve tasks collaboratively. Open full entry ’s assistant pattern, and the research agents at Anthropic and OpenAI all do thought-action- observation loops in different packaging.
Whether the thought-action labels need to be explicit in the output depends on the model. Reasoning models (o1, R1) internalize the thought step into their hidden chain-of-thought; non-reasoning models often benefit from the explicit ReAct prompting scaffold to keep the loop coherent.