Tool Calling
The concrete act of a model choosing a tool and issuing the request to run it, mid-response.
Think of it like
Reaching for the right implement on the workbench the moment the task needs it — grab the wrench, not just describe wrenches.
Example
Halfway through answering, the model emits a call to the `get_stock_price` tool with `ticker: "AAPL"` and pauses for the result.
How it actually works
Tool calling is the mechanical event inside tool use: the model outputs a structured request naming the tool and its arguments, the runtime pauses generation, runs it, and resumes with the result appended. Modern models are trained to do this natively, often calling several tools in one turn. The reliability question is whether it picks the right tool and fills the arguments correctly.
For product teams
The moment your model stops talking and starts doing — literally the trigger for action.
For engineers
The model emitting a structured tool invocation that suspends generation until the runtime returns a result.
Related
- Function Calling — The structured-output mechanism behind it.
- Tool Use — The umbrella idea it realizes.
- Tool Result — What the runtime hands back afterward.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome