DEV Community

Cover image for I Spent 4 Hours Fixing Broken Imports – So I Built a Complete Agentic AI Tutorial
zkaria gamal
zkaria gamal

Posted on

I Spent 4 Hours Fixing Broken Imports – So I Built a Complete Agentic AI Tutorial

*A true story from last month: *

I was building an intelligent agent using LangGraph + MCP, and I asked Claude for the latest code to implement a Multi-Node Agent.

It gave me a clean-looking code. I copied it, ran it… Import error.

I went to GPT-4o. Different code, but still outdated imports.

Tried Gemini. Same problem.

I lost over 4 hours tweaking imports, updating StateGraph, digging through LangChain's changelog… until I hit complete frustration.

That's when I said: Enough.

I decided to do something completely different.

I started collecting only the modern code that actually works in 2026, tested it myself, fixed what was broken, and organized everything in one place.

The result?

🔥 Agentic AI Tutorial

A complete, up-to-date reference for building Agentic AI using the latest versions of LangChain + LangGraph + MCP.


Why I built this repo

  • To end the daily struggle of "outdated/copied-pasted-broken code"
  • To help you build powerful agents without wasting hours on debugging
  • To give you one trusted, updated reference where everything actually runs

What's inside?

Chapter Content
1 LLM basics + Streaming + Advanced Prompts
2 LangChain LCEL + Tools + Chains
3 Advanced Memory + Full RAG (Chroma & FAISS)
4 Advanced LangGraph (ReAct, Router, Multi-Agent, Self-Refine, Human-in-the-Loop)
5 Complete MCP + FastMCP Server + Multi-Node Agent System (Router → Execution → Summary)

Everything is available as Jupyter Notebooks + Python files, ready to run.

Works locally (Ollama) and cloud (GPT-4o, Gemini).


A question for you:

Are you already at Chapter 5, or still at the beginning?

Drop a comment below 👇

AgenticAI#LangChain#LangGraph#MCP#Python#LLM#GenerativeAI#Opensource

Top comments (0)