---
published: true
id: tool-call
title: Tool Call
description: Knowledge captured in João Lopes' learning graph.
tags:
- ai-engineering
- agent-design
- llm
locale: en
resource: https://joaolopes.dev.br/knowledge/concepts/en/tool-call/
created: '2026-03-27'
---

# Tool Call
When the model asks the harness to do something — like read a file, run a command, or check a calendar. The model describes the action in text; the harness executes it.

## Definition
A tool call is how the model's text output becomes an action in the outside world. The model does not execute tool calls — it produces them as part of its response. The runtime reads the tool call, executes the tool, and feeds the result back as context for the next turn. A single response can contain zero, one, or multiple tool calls; zero means the loop ends, one or more means the runtime executes them and starts a new turn.

## Source
- [Tool use with Claude - Anthropic Docs](https://platform.claude.com/docs/en/agents-and-tools/tool-use/overview)

## Related
- [Tool](/concepts/en/tool.md)
- [Runtime](/concepts/en/runtime.md)
- [Turn](/concepts/en/turn.md)
