Defined term
ReAct
A reasoning pattern where the model alternates Thought → Action → Observation steps.
ReAct (Reason + Act) is a prompting pattern that interleaves chain-of-thought reasoning with tool calls. Each step: think about what to do, take an action (call a tool), observe the result, then think again. It is the basis of most modern agent loops. Production ReAct loops need timeouts, max-iteration caps, and budget controls to avoid runaway costs.
When it matters
When the workflow has more than 2 decision points and the model needs to inspect intermediate results. The foundational reasoning pattern for production agents.
Real example
A research agent using ReAct: Thought ('I need recent funding info'), Action (call_crunchbase_api), Observation ('Series B closed Mar 2026, $40M'), Thought ('Now check headcount'), Action (call_linkedin_api), Observation (...), Final Answer. Auditable trace, debuggable, recoverable.
KPIs to watch
Agent task completion rate (>80% on labelled tasks), step count distribution (most tasks <5 steps; outliers indicate prompt issues), debugging time per failure (target: <15 min with trace).
Related terms
Agentic AI
AI systems that can plan, take multi-step actions, and use tools to complete tasks autonomously.
Tool use
An LLM's ability to call external functions, APIs, or services within a generation step.
Chain of thought
Prompting the model to show intermediate reasoning steps before producing a final answer.
Autonomous agent
An AI agent that completes a defined task without per-step human input.
See it in action
We use this every week
Book a 30-min call and we'll walk you through how ReAct shows up in a real engagement we're running.
Book a 30-min call