Recipe Overview

LLMs lack direct access to real-time data or calculations. A tool-use agent integrates external APIs (search, calculators, databases) into the reasoning loop. The problem it solves is providing factual accuracy. When needed, the agent decides to call a tool. For example, for a weather query the agent might call a weather API instead of guessing. This solves hallucination by grounding responses in real data. The pattern enables agents to be both conversational and factually accurate, combining language understanding with access to current, verified information sources.