AI Agents with LangChain & LangGraph
Welcome, champion! 🔥
You've mastered 0 of 170 lessons. Start your first lesson free — no signup needed.
About this course
AI Agents with LangChain & LangGraph teaches you to build working AI agents in Python with the current LangChain 1.x and LangGraph 1.x APIs. The first lesson is a small agent you can run: create_agent with one @tool function, a question, the tool call it makes and the answer it returns. From there the course covers tools and tool errors, structured output with response_format, middleware for summarisation, PII redaction and human approval, LangGraph state graphs with reducers, conditional routing and Send, checkpointers, interrupts and time travel, short-term and long-term memory, retrieval as a tool, streaming, multi-agent supervisors and handoffs, LangSmith tracing and evaluation, and serving a graph as an API. The course has 170 lessons and about 1133 minutes of content. Each lesson is a short narrated reading card followed by MCQ practice with an explanation for every answer.
How the lessons work
Every lesson starts from a short piece of Python rather than a definition: an agent call and the messages it returns, a StateGraph with two nodes and a conditional edge, or a checkpointer resuming a paused run. Each idea is explained against that code, then MCQ practice asks what the code prints, which edge fires, or which change fixes a failing agent, with an explanation for every answer. Code uses LangChain 1.x and LangGraph 1.x only, so you learn create_agent and middleware rather than deprecated classes such as LLMChain or the old prebuilt ReAct agent.
Who this course is for
Python developers, students and freshers who have called an LLM API and now want to build agents that use tools, keep state and recover from errors. You should be comfortable writing Python functions, classes and type hints. Backend engineers adding an AI feature to a product, and data scientists moving from notebooks to agent workflows, will also find it useful. If you want the concepts behind agents across every framework, GenAI Engineering: Agents & Mastery covers them broadly; this course goes deep on the two libraries most teams build with.
MCQ practice vs video courses
Video courses usually walk through one demo agent from start to finish, which shows the happy path once. Real agent bugs are smaller and harder to spot: a reducer that overwrites messages instead of appending, a routing function that never reaches END, a tool error that ends the run, or an interrupt that resumes on the wrong thread. This course puts you in those situations with a few lines of code and asks what happens next, then explains why. That builds the judgement to debug a LangGraph agent instead of copying one.
Best way to learn LangChain and LangGraph
Keep a Python environment open and run each lesson's example yourself, changing one thing at a time: a tool's docstring, a reducer, an edge. Answer each MCQ before reading its explanation. Turn on LangSmith tracing early, because watching a run tree makes routing and tool calls visible. When you reach persistence and interrupts, build one small approval-gated agent end to end; that single project ties together most of the course.