What is tool use (function calling) in LLMs?
LLM & AILevel: FoundationTool Use & MCP
What is tool use (function calling) in LLMs?
Hints
- Identify the stage: pretraining, adaptation, generation or evaluation.
- Name the inputs, constraints and a case where the approach fails.
Answer
Function calling lets the model request execution of external tools by outputting structured JSON with tool name and parameters:
{
"tool": "get_weather",
"parameters": {"city": "Tokyo", "units": "celsius"}
}
The application executes the tool and returns the result. The model reasons over the result and either calls another tool or produces a final answer.
Modern LLMs (GPT-4, Claude, Gemini) are trained to generate syntactically correct function calls. You define available tools as a schema (name, description, parameter types) — the LLM selects the right tool and extracts parameters from the user's natural language request.
Source and licence: JustInCache & contributors — Awesome AI Interviews · Apache-2.0