loading
loading
The cycle where the model requests a tool, gets the result, and decides what to do next — repeating until done.
This is the engine of every agent: model emits a tool call, the runtime executes it, feeds the output back, and the model reasons again — looping until it produces a final answer instead of another call. The loop is where agency actually lives, and getting its stop condition and error handling right is most of the work in building reliable agents. Example: 'find the cheapest flight' becomes search → read results → search again with new dates → book, each turn driven by the previous tool's output.
Plainly
Think of Tool-calling Loop as a named building block in a big LEGO app. The cycle where the model requests a tool, gets the result, and decides what to do next — repeating until done.
In practice
Use it when you are mapping how the app is structured or explaining a feature to a teammate or agent. In practice, define the owner, input, output, and failure mode before you rely on it.