ReAct prompting
A framework where a model alternates reasoning traces and actions with tools or an external environment.
On this page
Source content checked . English translation published 2026-10-27.
In business automation
ReAct alternates reasoning and acting: a model chooses an action, receives an observation from the environment and continues until it can answer.
Examples
These examples illustrate a possible workflow. Adapt the connections and rules to your own systems.
Search question
The model decides to search, reads the observation and continues with the next reasoning step.
Relevant tools
- OpenAI Agents SDK
- Use ReAct as a prompting pattern to coordinate model reasoning with external tool actions.
What to check
- Unbounded tools can cause unwanted actions, and long action traces increase token use.