---
format: "aidr-story-markdown/v1"
id: "8fc272f28c8a0129df9cf25f82ba232387c9484d86476ab2e307267b043f8a2c"
canonical_url: "https://aidr.today/8fc272f2?lang=en"
title: "Plan or Retrieve First? A Scale Study of Tool Discovery Architectures"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-15T00:00:00.000Z"
category: "Research"
topics: ["google","agent"]
source_urls: ["https://machinelearning.apple.com/research/plan-retrieve-tool-discovery"]
summary: "A scale study of tool discovery architectures found that Retrieve-then-Plan (RtP) outperforms Plan-then-Retrieve (PtR) on single-tool queries, while PtR outperforms RtP on multi-tool queries, with the crossover consistent across retrieval budgets and model families."
---

# Plan or Retrieve First? A Scale Study of Tool Discovery Architectures

> [Open the canonical story](<https://aidr.today/8fc272f2?lang=en>)

**Published:** 2026-10-15T00:00:00.000Z
**Category:** Research
**Topics:** google, agent

## Summary

A scale study of tool discovery architectures found that Retrieve\-then\-Plan \(RtP\) outperforms Plan\-then\-Retrieve \(PtR\) on single\-tool queries, while PtR outperforms RtP on multi\-tool queries, with the crossover consistent across retrieval budgets and model families\.

## Sources

- [Story source](<https://machinelearning.apple.com/research/plan-retrieve-tool-discovery>)