Udemy — Agentic AI Internals: Build an Agent from Scratch 2026-3

Description

Eden Marco explores how an AI agent combines a language model, prompts, tools, state and a repeated decision-making loop. LangChain projects provide the practical setting for examining framework internals, navigating its source code and connecting components such as document loaders, text splitters, output parsers and memory.

The course relates prompting approaches including ReAct and few-shot examples to context management and multistep agent behavior. Retrieval-augmented generation, Pinecone and FAISS vector stores, MCP and LangGraph extend the discussion beyond a basic chat interaction. Workflow structure, error handling and evaluation are used to examine how an agent behaves across a sequence of tasks.

Intended audience: Software developers investigating agent internals and practical AI engineering.

What you will learn

• Explain the components and control flow of an agent application.
• Build LangChain workflows and inspect their implementation.
• Connect retrieval and vector-store components to agent context.
• Explore MCP, LangGraph, state handling and behavioral evaluation.

Course specifications

Publisher

Udemy

Instructors

Eden Marco

Language

English

Level

Beginner to Advanced

Duration

5 hours and 37 minutes

Lecture count

57

Quality

1080p

Edition

2026-3

Course topics

View course topics

Prerequisites

• Software-engineering familiarity with Python, Git, pipenv, environment variables, classes, testing and debugging.
• This is a nonbeginner course; prior machine-learning experience is not required.

Introduction video

Download resources

Choose one edition and keep its parts together. Alternate providers are available when listed.

Size

4.6 GB

Leave a Reply

Signed-in member comments appear immediately.

More to explore