What is Tool use?
Tool use is the capability that lets a language model request calls to functions defined by a developer or provided by the vendor, returning a structured call instead of plain text.
Also written: Claude tool use, LLM tool calling
Explained
Tool use lets a model such as Claude decide, based on the user's request and each tool's description, that it needs an outside function. Instead of answering in prose, the model returns a structured call naming the tool and its arguments. Anthropic's documentation says tool use is also called function calling.
The docs divide tools by where the code runs. Client tools run in your own application. That group covers tools you define yourself and tools with schemas published by Anthropic, such as bash and the text editor. For a client tool, the response ends with a stop reason of tool_use and one or more tool_use blocks. Your code performs the operation and sends the output back in a tool_result block, and the model then continues with that information. Server tools, such as web search, run on Anthropic's infrastructure, so the results arrive without any handler code on your side.
You define a custom tool with a name, a description and a JSON input schema. By default the model chooses on each turn whether to call a tool or answer directly, and the tool_choice setting can force a call. Tool definitions and tool blocks count as input and output tokens, so they add to request cost. The docs also note that when a required parameter is missing, a model may ask for it or may guess a value, so the behavior is not guaranteed.
Primary source
Tool use with Claude (platform.claude.com), read Oct 26, 2026
This entry paraphrases the page above. Sources change; check it for the current wording.
Listings that use this
The 5 top-ranked listings tagged for Tool use. Rank weighs adoption, freshness, documentation and publisher identity; how we rank and grade. Money never changes the order.
- 0.791score
ADK for Go
Google
Code-first Go toolkit for building agents with a tool ecosystem, composable multi-agent systems and Cloud Run deployment.
HarnessAFreshness: A★ 8.9K - 0.753score
LangChain
LangChain
Python framework of interoperable components and integrations for building agents and LLM apps, with LangGraph and Deep Agents layered above it.
HarnessAFreshness: A★ 148.4K↓ 37.3M/wk - 0.734score
Pydantic AI
Pydantic
Typed Python SDK for building agents with swappable model providers, plus optional capabilities for memory, sub-agents, planning and durable execution.
HarnessAFreshness: A★ 20.8K↓ 1.5M/wk - 0.734score
Google Agent Development Kit (ADK)
Google
Open-source toolkit for building, evaluating and deploying agents in Python, TypeScript, Go, Java and Kotlin, tuned for Gemini but model-agnostic.
HarnessAFreshness: A★ 22K↓ 2.4M/wk - 0.730score
Claude Agent SDK (Python)
Anthropic
Python SDK that runs the Claude Code agent loop inside your own program, with custom tools and hooks written as Python functions.
HarnessAFreshness: A★ 8.3K↓ 7.9M/wk
Related
Cite this page
What is Tool use?. Agent Mart. https://agentmart.dev/glossary/tool-use (last updated 2026-10-26)
Written Oct 26, 2026, updated Oct 26, 2026. All glossary terms