Given a set of tool definitions, the model can respond with a structured call — a function name and arguments matching a schema — instead of plain text. Your code executes it and returns the result, which the model uses to continue. This loop is what turns a chatbot into an agent that can act.
How reliably a model and framework handle tool calling is a major differentiator between agent stacks. Open standards like the Model Context Protocol standardise the tools themselves so they are reusable across clients.