Build and Evaluate AI Agents with LangChain and LangSmith

About this course

This course contains the use of artificial intelligence.

You know some Python and you have used ChatGPT. Now you want to build an assistant that looks things up, takes actions and does not make things up. The LangChain tutorials you find are often written for older versions, or they jump straight to a finished agent, so when something breaks you cannot tell which piece failed.

This course starts from zero and builds one assistant, piece by piece. Our example is Bloom, a small made-up plant shop with real data: 30 plant pages, an orders database and a return policy. In every lesson the Bloom assistant gets one new ability, and you see exactly what changed.

Every lesson has three parts:

An explainer with diagrams and animations: the key words first, then the idea block by block, worked on Bloom data by hand before any code. A live-code video: the lesson notebook, cell by cell, with its real outputs. You type along. A quiz and a hands-on Udemy Lab: short exercises in a slightly different situation, with checks that print ✓ when your code is right, and a worked solution. What makes it different

Current APIs. LangChain 1.x with create_agent and middleware, LangGraph 1.x, the LangSmith SDK, MCP adapters and Deep Agents, versions pinned so the code keeps running. Real runs, real numbers. Every answer, token count and cost on screen comes from a real model call. When the model invents a tracking number or calls the wrong tool, the course shows it, finds the cause in the trace, and fixes it. Free to follow. Examples run on the open-weight gpt-oss model through Groq's free tier, or on your own computer with Ollama. Setup shown on screen.

Read the full description on Udemy on Udemy (opens in a new tab)

Prices, discounts and availability are set by Udemy. We may earn a commission when you purchase through links on this site.