In this video, I explain how tool calling works with LLMs.
Tool Calling is a fundamental concept for understanding AI Agents. LLMs like gpt-4o and Claude Sonnet are unable to execute code or do anything other output tokens. AI Agents need a way to interact with real world systems, and tool calling is the way that this is accomplished.
In this video, I explain how tool calling works at a high level. I walk through a series of prompts that showcase the main concepts of tool calling, and then I give some code examples in typescript. I also explain how tool calling fits into MCP (Model Context Protocol) and how to hook up an MCP Server's tools into an OpenAI API request.
Tool calling is not nearly as scary as it seems - I hope you all enjoy the video!