AI agent glossary
ReAct
A pattern where a model alternates between reasoning about what to do and acting with a tool, using each observation to guide the next step.
ReAct, short for reasoning and acting, comes from a 2022 research paper. The idea is to have the model interleave thoughts with actions: think about the situation, take an action such as a search, observe the result, then think again.
This was an influential step toward modern agents. Reasoning alone can drift away from the facts, and acting alone has no plan. Combining them lets the model adjust its approach as evidence comes in.
Today most agent loops follow the ReAct shape naturally, using native tool calling rather than the text-parsing format from the original paper. Knowing the name helps when reading older tutorials and frameworks.
Thought: I need the population of the city. Action: search for it. Observation: the search returns a figure. Thought: now I can compare it with the other city.
Learn more in the course
Related terms
Go from definitions to doing
Learn to use and build AI agents in 65 practical lessons.