Learn
What is tool calling?
Last reviewed July 16, 2026
Tool calling (also called function calling) is the mechanism that lets a language model act: you describe available tools with typed parameters, the model replies with a structured call (a tool name plus JSON arguments) instead of prose, your code executes it, and the result returns to the model to continue. Every agent loop is tool calls repeated toward a goal.
How a tool call round-trips
Four steps: the application advertises tools with names, descriptions, and JSON-schema parameters; the model outputs a call; the application validates and executes it (the model never runs code itself); and the result is appended to the conversation for the next model turn. The schema is the contract: good parameter descriptions do more for reliability than prompt exhortations.
Reliability and safety
Models sometimes call the wrong tool, malform arguments, or loop; production systems validate arguments, bound retries, and meter budgets. Because the application executes every call, it is also the enforcement point: permissions, confirmation gates on destructive actions, and audit trails belong there, not in the prompt.
idapt agents address every capability (files, web search, code, computers) through one structured tool interface, with per-tool permissions, confirmation gates, and a full call trace per run.
Frequently asked
See it in practice
The fastest way to understand it is to use it. Start free, no credit card.
Try idapt free