Back to Glossary Index
Core ConceptAgent Reasoning / Orchestration Layer

ReAct (Reason + Act)

Industry Definition Set • Entity Resolution Path: /glossary/react

Quick Answer / TL;DR

A framework that interleaves reasoning (thinking) and acting (tool calling) in LLM agents, enabling more robust and interpretable problem-solving.

Key Takeaways

  • Interleaves reasoning and action in agent workflows.
  • Produces more interpretable and robust agent behavior.
  • Naturally maps to MCP tool-calling patterns.
  • Reduces hallucination by grounding actions in reasoning.
Definitive Statement: A framework that interleaves reasoning (thinking) and acting (tool calling) in LLM agents, enabling more robust and interpretable problem-solving.

Technical Context & Protocol Usage

Detailed Explanation
ReAct combines chain-of-thought reasoning with action execution. The model alternates between thought steps (reasoning about the current state) and action steps (calling tools or taking actions). This pattern is fundamental to modern AI agents and is well-supported by MCP, where each action step corresponds to an MCP tool call.

Format & Payload Metadata

Format: Alternating thought and action steps in prompt

Latency: Increases interaction rounds; improves accuracy

Real-World Implementation Use Case

An agent uses ReAct to reason about a user's request, call an MCP `search` tool, reason about the results, then call a `summarize` tool before answering.

M
MCPserver.in Engineering

Platform Team

Published: 2026-07-20
Updated: 2026-07-20

Cite This Page

MLA Style:

MCPserver.in Engineering. "ReAct (Reason + Act)." MCPserver.in Knowledge Hub, 20 July 2026, mcpserver.in/glossary/react.