Tool Use
The ability of a model to invoke external tools — APIs, code executors, search — to augment its capabilities.
Full Definition
Tool use (Anthropic's term for what OpenAI calls function calling) allows a language model to decide when to invoke external capabilities — web search, code execution, database queries, file I/O, or arbitrary APIs — instead of relying on its parametric knowledge. The model outputs a structured tool call; the application executes it and returns results; the model continues reasoning with those results. This turns the model from a static knowledge retriever into a dynamic agent capable of taking actions in the world. Tool use is foundational to agentic systems and dramatically expands what LLMs can do while reducing hallucination on tasks requiring current data or computation.
Examples
Claude calling a Python interpreter tool to execute code, then using the output to answer a numerical question accurately.
An agent using a web_search tool to look up a stock price, then a database_query tool to pull the user's portfolio, then synthesising both into a portfolio analysis.
Apply this in your prompts
PromptITIN automatically uses techniques like Tool Use to build better prompts for you.