This is the BEST Way to build Agentic AI Apps скачать в хорошем качестве

This is the BEST Way to build Agentic AI Apps

Most AI agents look impressive in demos, but fall apart the moment you try to ship them. In this video, I break down the architectural mistakes I see developers make when building agentic applications, especially around tool calling, authentication, and execution. We’ll look at why the typical Next.js + LLM setup isn’t enough, and what actually changes when you want agents to take real actions in production. I’ll also walk through a working example and the infrastructure pattern that makes it scalable. Enjoy! 🔥 Try out Arcade for FREE: https://arcade.dev.plug.dev/GUvDlgK 📸 Screen Recording Software: https://dub.sh/eDa47SO 🔒 My favorite Auth Solution: https://dub.sh/xeU8r3v ⌨️ GitHub Repo: https://github.com/ski043/nextjs-fina... 🌍 My Website: https://www.syntaxpath.com/ 👋🏻 Discord:   / discord   ✅ Follow me on X: https://x.com/janmarshaldev 📧 Business ONLY: jan@alenix.de Timestamps: 00:00 Intro 01:08 The Fundamental Mistake Most People Make 07:02 Tool Calls Are Not Scalable (Orchestration) 09:20 MCP Is Only Part of the Solution (OAuth, Permissions, Token Refresh, etc.) 14:50 Agentic Auth Is Essential 18:00 High-Level Architecture Overview 22:00 Agent Approach in Code 29:00 Request Flow (AI Agent, Engine, Tool, API) 36:00 Going from Pipeline to Agent 38:00 Custom MCP Server (Essential)