mediumAI Engineering

How does function calling (tool use) work under the hood?

Q: How does function calling (tool use) work under the hood?

**TL;DR.** You send **tool definitions** (name, description, JSON-Schema params) with the conversation. The model **does not execute anything** — it returns a structured **tool-call message** (name + arguments); your code runs the function, appends the **result message**, and calls the model again until it answers in text. The model plans; **your runtime executes**. model emits tool_use: name+args <text class='d-text' x='35

127 views

Understand the problem

Schemas in, structured intent out: the request-response loop that lets models act.

function-callingtool-useagents

Attempt it yourself

Sketch your approach before reading the solution — that's what interviews test.

Stuck? AI Nudge Available

Get a conceptual hint to guide your logic without spoiling the final implementation.

Study the solution

The solution is waiting

Give it an honest attempt first — then compare your thinking with the full walkthrough.

Read the code

The canonical tool loop

let messages: Msg[] = [{ role: "user", content: task }];

for (let step = 0; step < MAX_STEPS; step++) {
  const res = await client.messages.create({ model, tools, messages, max_tokens: 1024 });
  if (res.stop_reason !== "tool_use") return textOf(res);       // done

  const calls = res.content.filter((b) => b.type === "tool_use");
  const results = await Promise.all(calls.map(async (c) => ({
    type: "tool_result", tool_use_id: c.id,
    content: await execValidated(c.name, c.input),               // schema-check first
  })));
  messages.push({ role: "assistant", content: res.content },
                 { role: "user", content: results });
}
throw new Error("step budget exhausted");

Join the discussion

Discussion (0)

No responses yet. Be the first to share what you think.

How do you catch regressions when changing a prompt or upgrading a model?

What is the ReAct pattern and why did it become the default for agents?