ReWOO

The agent plans all necessary tool calls upfront, executes them in batch, and uses the aggregated results for the final answer. Typical conditions for use: The required tool calls are identifiable at the beginning; LLM calls need to be drastically reduced. The central trade-off: Significantly lower LLM costs and latency, but at the expense of adaptability du…

Use this profile to understand the building block briefly, place it in the model, and switch to the 360° assessment when needed.

Content type

Concept

Theoretical construct: explains a term, principle, or mental model.

Classification level

Structure

What organizes, connects, or makes decisions possible.

Available perspectives

Open this building block directly in an already curated view.

360°

In brief

ReWOO in brief

ReWOO is a language-model planning approach that plans tool calls first and executes them afterward as a batch.

Where does it come from?

Xu and colleagues introduced the ReWOO framework in 2023. It arose from the problem that classic ReAct loops tightly interleave planning, tool calls, and renewed reasoning, which can be costly and error-prone.

The mental model

A travel agency first creates the complete itinerary with placeholders for booking results. Only then are the bookings executed and their results inserted into the plan. Planning does not have to restart after every call.

Key building blocks

Plan

A sequence of planned steps with placeholders for later results.

Tool execution

Executing the tools specified by the plan separately from planning.

Result variable

A placeholder referring to the result of an earlier step.

When is it useful?

ReWOO can make tool chains more efficient when steps are predictable. New information or a flawed plan requires adaptation and can make it less flexible than a stepwise loop.

Xu et al., ReWOO: Decoupling Reasoning from Observations for Efficient Augmented Language Models