---
schema_version: "newruntime-agent-readable-v0.2"
type: "post"
stable_id: "post:anthropic-tool-search-programmatic-calls"
slug: "anthropic-tool-search-programmatic-calls"
title: "Anthropic Moves Large Tool Libraries Out Of Context"
description: "Anthropic's Tool Search Tool, Programmatic Tool Calling, and Tool Use Examples separate discovery, orchestration, and usage guidance for agents with large tool libraries."
retrieval_nugget: "Anthropic's Tool Search Tool, Programmatic Tool Calling, and Tool Use Examples separate discovery, orchestration, and usage guidance for agents with large tool libraries. Anthropic's advanced tool-use release addresses three separate failure modes that appear when an agent connects to dozens of systems. First, tool definitions consume context before work begins."
status: "published"
published_at: "2026-08-01"
updated_at: "2026-08-01"
record_date: "2026-08-01"
date_kind: "published_at"
topics: ["tools","context-engineering","mcp","agent-harnesses"]
source_urls: ["https://www.anthropic.com/engineering/advanced-tool-use"]
visuals: [{"id":"anthropic-tool-search-programmatic-calls","kind":"editorial-diagram","role":"hero","src":"https://newruntime.com/images/posts/anthropic-tool-search-programmatic-calls.webp","alt":"Hand-drawn tool architecture where a compact search index reveals a few relevant tools, code orchestrates them, and only final results enter model context.","caption":"Anthropic splits large-tool agent work into on-demand discovery, code-based orchestration, and example-backed usage.","credit":"New Runtime synthesis from Anthropic Engineering","source_url":"https://www.anthropic.com/engineering/advanced-tool-use","generated_with":"gemini-3.1-flash-image","width":1600,"height":900,"legend":[{"label":"Discover","description":"The agent searches a compact index and expands only the relevant tool definitions."},{"label":"Orchestrate","description":"Code handles loops, conditions, transformations, and intermediate results outside model context."},{"label":"Demonstrate","description":"Examples teach valid usage patterns that a JSON schema cannot express."}]}]
telegram_message_id: 2894
telegram_url: "https://t.me/qwgai/2894"
telegram_message_ids: [2894,2895]
telegram_delivery_mode: "text_then_media"
telegram_media_url: "https://t.me/qwgai/2895"
routes: {"html":"https://newruntime.com/posts/anthropic-tool-search-programmatic-calls/","markdown":"https://newruntime.com/posts/anthropic-tool-search-programmatic-calls.md","json":"https://newruntime.com/posts/anthropic-tool-search-programmatic-calls.json"}
source_format: "markdown"
---

# Anthropic Moves Large Tool Libraries Out Of Context

## Retrieval answer

Anthropic's Tool Search Tool, Programmatic Tool Calling, and Tool Use Examples separate discovery, orchestration, and usage guidance for agents with large tool libraries. Anthropic's advanced tool-use release addresses three separate failure modes that appear when an agent connects to dozens of systems. First, tool definitions consume context before work begins.

Anthropic's advanced tool-use release addresses three separate failure modes that appear when an agent connects to dozens of systems.

First, tool definitions consume context before work begins. Anthropic's example combines GitHub, Slack, Sentry, Grafana, and Splunk into 58 tools and roughly 55,000 tokens of definitions. The company says it has observed 134,000 tokens of tool definitions in larger setups. Similar names then make selection harder, not easier.

The Tool Search Tool changes the loading model. Developers can mark tools or whole MCP servers with `defer_loading: true`. Claude starts with the search capability and a few essential tools, then expands only the definitions needed for the task. Anthropic reports an 85% reduction in tool-definition token use in one internal setup, alongside better MCP evaluation accuracy for Opus models.

Second, Programmatic Tool Calling lets Claude express orchestration as code. Loops, conditions, joins, and data transformations run in a code-execution environment. Large intermediate tool results are processed there, and only the compact output returns to model context. That avoids a full model round trip for every mechanical step.

Third, Tool Use Examples cover semantics that schemas miss. A JSON schema can say an optional field is valid. It cannot reliably teach when the field should be used, which parameter combinations are conventional, or what an API expects in a real workflow.

The combined pattern is stronger than adding a larger context window. Keep discovery cheap, orchestration explicit, and examples close to the tool contract. Load tokens only when the current task proves they are necessary.