Tool calling
도구 호출
A setup where an AI asks, in a fixed format, for an outside tool such as search or a calculator, then uses the result to answer.
Overview
A language model only reads and writes text. On its own it cannot search, calculate or use files. So you tell it in advance which tools it may use, and the model writes a request in a fixed format: "call this tool with these values."
Then a program outside the model runs the tool for it and sends the result back. The model reads the result and either writes an answer or calls the next tool. An agent can finish a multi-step job because it repeats this loop.
An everyday example
It is like a chef who never leaves the kitchen but writes a note, "Bring me ten eggs from the storeroom," and an assistant fetches them.
Why it matters
It is the basic technology that turns AI from a machine that only talks into a machine that does work. Models differ in how accurately they call tools, so people compare this separately when choosing a model for an agent.
Also calledfunction calling