What is ReAct, and why was it influential?

beginner

Re-explain

ReAct, short for Reasoning and Acting, is a prompting pattern from a 2022 paper by Yao et al. It makes one model interleave short reasoning steps with tool actions in a single trace.

Thought: I need Acme's current CEO.
Action: search("Acme CEO")
Observation: Jane Doe, since 2024.
Thought: That answers it.
Answer: Jane Doe.

It was influential because it joined two things that were separate. Chain-of-thought, where the model writes its reasoning out loud, could plan but not check facts. Tool-using systems could act but not plan. ReAct grounds each reasoning step in a real observation, so wrong turns get caught mid-task.

Today most agents run this shape through native tool-calling APIs instead of parsing Thought: and Action: text.

Rewriting in plainer words…

This answer doesn't lend itself to a diagram - it reads best . No credits were charged.

Why there's no diagram: “”

The interactive diagram is below the answer - jump to diagram ↓ · Below it, the related concept . Jump to it ↓

The diagram below the answer is the concept . Jump to it ↓

Tailored explanation · switch back to · ·

What should the new diagram focus on?

Guided practice for AI Agents One question at a time. Answer out loud, get graded, see what you missed.

How well did you know this?

AI:

Saved in this browser - sign in to keep your review list.

How should your speech become text?

Listening… your words appear above as you speak - tap Stop when you're done.

Recording · cr - tap Stop & transcribe when you're done.

Transcribing with AI…

Voice:

Keep going - a few more words and AI can grade it.

Interview lens How interviewers actually use this question

Likely follow-ups

Why do modern agents use native tool-calling APIs instead of parsing Thought: and Action: text? Name what text parsing breaks on, like malformed actions or the model writing its own Observation, and what structured tool calls fix.
When would you choose plan-then-execute over ReAct-style interleaving? Weigh adaptivity against cost. ReAct pays one model call per step and adapts. A plan made upfront is cheaper and easier to audit, but it is brittle.
What does your ReAct loop do when a tool call fails or returns something useless? Feed the error back as an observation the model can read, cap retries, and stop or escalate instead of retrying blindly.
How do you keep a ReAct agent from looping forever? Step and token budgets, spotting the same action repeated, and a clear stop condition or handoff to a human.

What you can say

ReAct, short for Reasoning and Acting, is a prompting pattern where one model alternates short reasoning steps with tool calls in a single trace.
It comes from a 2022 paper by Yao et al., and the trace reads Thought, Action, Observation, repeated until the model can answer.
Say the model needs a company's current CEO. It notes that, calls search, reads the result, and answers from what it saw.
It mattered because it joined two separate lines of work. Chain-of-thought could plan but not check facts, and tool-using systems could act but not plan.
Each reasoning step rests on a real observation, so the model catches a wrong turn in the middle of the task.
The cost is a model call per step, and today most agents run this same loop through native tool-calling APIs rather than parsing Thought and Action text.

Weak answers to avoid

Only expands the acronym Saying 'reasoning and acting' shows you read a glossary. Walk one Thought, Action, Observation cycle and say what it fixed that chain-of-thought could not.
Treats ReAct as a framework or library ReAct is a prompting pattern from a paper, not a package. Frameworks implement the loop, so describe it as the pattern they run.
Skips why it was influential Half the question is the why. Say it joined planning with acting, so real observations catch wrong turns mid-task.
Implies agents still parse Thought/Action text Describing production agents that regex Thought: and Action: lines sounds dated. Say the loop survives but now runs on native tool-calling APIs.

Interview lens

Likely follow-ups, what you can say, and the weak answers to avoid.

Want a quick review of the fundamentals? See the AI Agents cheatsheet.

← Back to all AI Agents questions

Pro · $10/mo

47 of 54 AI Agents answers are in Pro.

Full answers, code samples, and AI explanations that go simpler or deeper. Cancel anytime.

Full answers + code
AI explanations, simpler or deeper
1,000 AI credits / month
Cancel anytime

Get Pro → See plans & pricing →