What is Tool calling?
Tool calling (also called function calling) lets a model request that your code run a named function with arguments, then continue with the result — the foundation of AI agents.
With tool calling, you describe the functions your app exposes (name, description, JSON schema for arguments). The model can then return a structured request to call one of them instead of answering directly. Your code runs the function, returns the output, and the model uses it to produce the final answer.
This is what turns a chat model into an agent: it can look up data, call APIs, run code, or take actions, deciding on its own which tools to use and when. Most modern instruction-tuned models support it; some smaller or specialized models do not.
How Kyma does it
Kyma exposes a supportsTools flag per model and, for tool-using requests, routes only to tool-capable models — including the fallback chain — so an agent request never lands on a model that can't call functions.