ReAct Pattern
An agent control loop that interleaves Reasoning and Acting steps — think, act, observe, repeat.
What is ReAct Pattern?
The ReAct pattern (Reason + Act) structures an agent's execution as alternating reasoning traces and tool calls, with each observation feeding the next thought. It is the most widely adopted agent loop because it is inspectable: you can read why the agent chose each action. See ReAct Agent Pattern and Agentic Design Patterns.
What is an example of ReAct Pattern?
To answer an inventory question, an agent identifies the needed warehouse, calls the inventory tool, observes that the SKU is discontinued, queries the replacement catalog, and then returns the supported answer.
How is ReAct Pattern different from related concepts?
Chain-of-thought prompting focuses on intermediate reasoning. ReAct adds actions and observations from an external environment, creating an iterative agent loop.
What should enterprises evaluate for ReAct Pattern?
- Use typed tools and validate arguments and results at every action boundary.
- Set step, cost, time, and retry limits, with explicit stop and escalation conditions.
- Evaluate task success, tool choice, evidence use, recovery from bad observations, and unsafe-action attempts.
Read the full guide: ReAct Pattern — in-depth article →
Related terms
Authoritative sources
Primary sources for the formal meaning, requirements, or original research behind ReAct Pattern:
Putting ReAct Pattern to work?
VDF AI runs governed AI agents on your own infrastructure — on-premises, sovereign cloud, or air-gapped. Book a working session to map the architecture.
Talk to VDF AI