Function Calling is a capability that lets an AI model use external tools — such as searching the web, running code, or querying a database — by producing structured requests that a program can execute.
What it means in plain English
On its own, a language model can only generate text. Function calling extends it: given a set of available tools, the model can decide to “call” one, outputting a structured request (like “get_weather(city=London)”) that the surrounding software runs, returning the result for the model to use. This lets AI go beyond its training to fetch live data, take actions, and connect to other systems.
It is a foundational building block for AI agents.
A simple example
Ask an AI assistant “what’s the weather in Paris?” and, with function calling, instead of guessing, it calls a real weather tool, gets the current data, and answers accurately — combining its language ability with live, factual information.
Why it matters
Function calling connects language models to the real world of live data and actions, and is essential for building capable AI agents and assistants that do more than talk. It is a key mechanism turning AI from a text generator into a system that gets things done.
Related terms
- Agent — relies on function calling to use tools.
- API — what functions often connect to.
- Grounding — function calling can ground answers in real data.
Frequently asked questions
What is function calling in AI?
It lets a language model output a structured request to run a specific function or tool (like a calculator, search, or database query), so it can take actions and use external capabilities reliably.
Why is function calling important for agents?
It is a key mechanism that lets AI agents connect to real tools and APIs, turning a text model into something that can actually do things.
