AI Agents in LangGraph
by DeepLearning.AI
Build a ReAct agent by hand, then rebuild it in LangGraph — taught by LangChain's own CEO.
Overview
AI Agents in LangGraph is a free DeepLearning.AI short course taught by Harrison Chase, co-founder and CEO of LangChain, with Rotem Weiss, co-founder and CEO of Tavily — so it comes from the people who maintain both the framework and the agentic-search API it uses. It runs 1 hour 42 minutes across nine video lessons with six runnable notebooks and one graded quiz: Introduction, Build an Agent from Scratch, LangGraph Components, Agentic Search Tools, Persistence and Streaming, Human in the loop, Essay Writer, LangChain Resources, and Conclusion. The pedagogical structure is build-then-rebuild, and that is the reason to take this one over the many LangGraph tutorials that skip straight to the framework: you first implement a ReAct-style agent by hand around an LLM's tool-calling loop, so you can see the prompt, the parser and the control flow with nothing hidden, and only then rebuild the identical agent with LangGraph's StateGraph, nodes and conditional edges — which makes it obvious what the framework is actually buying you. Later lessons add checkpointers so agent state persists across threads, streaming at both step and token granularity, and interrupts for human approval before a tool fires. The capstone essay writer models a researcher's workflow: plan, research via Tavily's agentic search, draft, critique, revise. The stated prerequisite is intermediate Python. LangGraph itself is a low-level orchestration framework from LangChain Inc. that mixes deterministic hand-coded steps with LLM-driven ones in the same graph, usable without LangChain, and in production at Klarna, Uber and J.P. Morgan. Note that the framework has moved on since the course was recorded and its docs now live at docs.langchain.com — the concepts hold, but check current signatures there before copying code into a project.
At a Glance
- Topic
- Agentic
- Level
- Intermediate
- Format
- Course
- Cost
- Free
- Duration
- 1 hour 42 minutes, self-paced — 9 video lessons, 6 code notebooks, 1 graded quiz
- Provider
- DeepLearning.AI
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Implement a ReAct-style agent loop from scratch in plain Python before touching any framework
- ✓Rebuild that same agent with LangGraph StateGraph nodes, edges and conditional routing
- ✓Add checkpointers so agent state persists across separate threads and sessions
- ✓Stream agent output at both the step level and the individual token level
- ✓Interrupt a running graph for human approval before a tool call is allowed to execute
- ✓Use agentic search that returns multiple answers in a predictable, structured format
- ✓Compose a multi-step essay writer that plans, researches, drafts, critiques and revises
Highlights
- •Taught by LangGraph's own maintainer — Harrison Chase, co-founder and CEO of LangChain
- •Build-from-scratch first, framework second, so you can see exactly what LangGraph abstracts away
- •Under two hours end to end, with six runnable notebooks rather than slideware
- •Covers the production concerns most agent tutorials skip entirely: persistence, streaming, and human-in-the-loop gates
- •Free to watch and run, with the graded quiz the only piece behind a membership tier
- •The framework it teaches is in production at Klarna, Uber and J.P. Morgan, not a toy
Who It’s For
Best For
- ✓Python developers building their first stateful, multi-step LLM agent
- ✓Engineers deciding whether a graph framework beats a hand-rolled agent loop
- ✓Teams that need human approval gates inside an otherwise autonomous workflow
Prerequisites
- •Intermediate Python, including classes, decorators and the basics of async
- •Familiarity with calling an LLM API and with function/tool calling
- •An OpenAI and a Tavily API key if you want to run the notebooks outside the hosted environment
FAQ
What is AI Agents in LangGraph?
A free DeepLearning.AI short course taught by Harrison Chase of LangChain and Rotem Weiss of Tavily. In under two hours you implement an agent loop from scratch in plain Python, rebuild it with LangGraph's graph primitives, then add persistence, streaming and human-in-the-loop interrupts. The capstone is an essay-writing agent that plans, researches, drafts, critiques and revises like a human researcher would.
Is AI Agents in LangGraph free?
AI Agents in LangGraph is free to access.
What level is AI Agents in LangGraph for?
AI Agents in LangGraph is aimed at a intermediate audience. Recommended background: Intermediate Python, including classes, decorators and the basics of async, Familiarity with calling an LLM API and with function/tool calling, An OpenAI and a Tavily API key if you want to run the notebooks outside the hosted environment.
How long does AI Agents in LangGraph take?
Expect roughly 1 hour 42 minutes, self-paced — 9 video lessons, 6 code notebooks, 1 graded quiz. Most learners work through it at their own pace.
What will I learn from AI Agents in LangGraph?
You'll learn: Implement a ReAct-style agent loop from scratch in plain Python before touching any framework; Rebuild that same agent with LangGraph StateGraph nodes, edges and conditional routing; Add checkpointers so agent state persists across separate threads and sessions; Stream agent output at both the step level and the individual token level; Interrupt a running graph for human approval before a tool call is allowed to execute; Use agentic search that returns multiple answers in a predictable, structured format; Compose a multi-step essay writer that plans, researches, drafts, critiques and revises.
Topics
Sources
This page was written from 2 sources, 1 on domains other than deeplearning.ai.