agentic tool calls

Agentic tool calls enable LLMs to autonomously select, parameterize, and execute external functions to bridge the gap between reasoning and action.

Modern models like GPT-4o and Claude 3.5 Sonnet use agentic tool calling to move beyond static text generation. By defining specific JSON schemas for functions (e.g., get_weather or query_database), developers allow the model to pause execution, request a tool call, and process the resulting data. This loop transforms a chatbot into an active agent capable of interacting with APIs, legacy databases, and local file systems. Frameworks like LangGraph and CrewAI leverage this capability to build multi-turn workflows where the model decides which tool to use based on the real-time state of the conversation.

https://python.langchain.com/docs/concepts/tool_calling