Tool Use

Why Tool Use Matters

On their own, LLMs are text-in, text-out systems. They cannot browse the internet, run calculations precisely, read a file, or send an email. They are extraordinarily good at language and reasoning, but they have no hands.

Tool use (also called function calling) gives them hands.

How It Works

1. You define a set of tools, functions the model is allowed to call, by describing them in plain language and specifying their parameters.
2. When the model decides it needs to use a tool, it outputs a structured request instead of a direct answer: "call search_web with query = 'latest AI news'".
3. Your application executes the actual function and feeds the result back to the model.
4. The model incorporates the result and continues its response, or decides to call another tool.

The model never executes code directly. It requests; your system executes.

Common Tools in the Wild

• Web search: retrieve up-to-date information
• Code interpreter: run Python, do math, process data
• Database queries: look up customer records, product inventory
• Calendar and email: schedule meetings, send messages
• File read/write: read a document, save outputs

Tool Use is the Bridge to Agency

A model with tool use can accomplish multi-step tasks in the real world. This is why tool use is the foundational capability that separates a basic chatbot from a true AI agent. More tools means more real-world reach, and more responsibility for the developer to define what the model is allowed to touch.