0PricingLogin
AI Agents · Lesson

ReAct: Reason + Act Pattern

Thought -> Action -> Observation -> Thought ... — the loop that powers most modern multi-step agents.

What Is ReAct?

ReAct stands for Reason + Act. It is the canonical pattern for multi-step LLM agents — published by Yao et al. in 2022 and still the foundation of most modern agents.

The Loop

At each step, the agent emits three things:

  1. Thought — what to do next, in natural language
  2. Action — which tool to call and with what arguments
  3. Observation — the tool result, fed back in

Repeat until the agent emits Final Answer.

All lessons in this course

  1. ReAct: Reason + Act Pattern
  2. Implementing ReAct from Scratch
  3. Common Tool Sets (Web, Calculator, RAG)
  4. Detecting and Recovering from Tool Errors
← Back to AI Agents