AI & Automation
LangGraph: Build AI Agents That Think, Act & Collaborate
Design, build and debug reliable AI agents — from routing and tool calling to memory, human approval steps and multi-agent teams.
- Beginner to Intermediate
- 2 days
- HRD Corp claimable
Overview
LangGraph is one of today’s leading frameworks for building AI agents that go beyond simple chatbots — agents that route work, call tools, remember context, pause for human approval and collaborate with other agents.
In this 2-day hands-on training, participants learn LangGraph’s core concepts and build real agents from scratch, with a lab in every session. Day 1 covers the fundamentals and a first working agent; Day 2 moves to advanced patterns, human-in-the-loop and multi-agent systems, and finishes with a capstone project participants take home.
The course is for software developers and engineers exploring AI agents, data and AI practitioners who want to move beyond simple prompts, and tech leads and solution architects evaluating agent frameworks. You need basic Python (functions, classes and dictionaries); some familiarity with ChatGPT-style tools helps but is not required. Taking LangChain first gives a good foundation, but it is not required.
Participants leave with their capstone agent, combining routing, tools, memory and a human approval step, plus a Certificate of Completion. Join a public session, or run it in-house as a private cohort, on-site or online, built around your team’s use cases. It is HRD Corp claimable, subject to grant approval; see how HRD Corp claims work.
After the course
An agent that works in a lab still has to be planned, deployed and maintained in your own system. Docker Essentials covers packaging and deployment, and Python Intermediate with Ollama shows how to run models on your own machines.
Schedule is tentative and subject to change based on the trainer’s delivery, participants’ progress, and training needs.
What you’ll learn
- Explain what an agentic workflow is and when to use one
- Build LangGraph workflows with nodes, edges, state and reducers
- Route workflows with conditional edges and build loops that stop safely
- Connect agents to external tools and handle tool errors gracefully
- Build a ReAct agent from scratch and produce structured outputs
- Give agents memory and resume workflows with checkpointers
- Add human-in-the-loop approval before sensitive actions
- Build supervisor-led multi-agent systems with subgraphs
- Debug and trace agents with LangGraph Studio and LangSmith
Who should join
- Software developers and engineers exploring AI agents
- Data and AI practitioners who want to move beyond simple LLM prompts
- Tech leads and solution architects evaluating agent frameworks
Prerequisites: Basic Python (functions, classes, dictionaries). Some familiarity with LLMs or ChatGPT-style tools helps but isn't required. Bring a laptop with Python 3.10+ and an LLM API key — a setup guide is sent before the training.
Course modules
-
01 Introduction to LangGraph & Agentic Workflows
What AI agents are, how chains differ from agents, why graph-based workflows help, where LangGraph fits in the LangChain ecosystem, and business problems that suit it.
-
02 Core Concepts: Nodes, Edges & State
Defining graph state with TypedDict, nodes as Python functions, edges, START and END, reducers (such as appending messages), and compiling and running a graph.
-
03 Conditional Routing
Conditional edges, routing functions, branching, loops and exit conditions, applied to classification and triage. Lab: build a ticket triage workflow.
-
04 Tool Calling
Defining tools, binding them to an LLM, ToolNode, handling results and errors, and how an LLM decides when to call a tool.
-
05 Building Multi-Step Agents
The ReAct pattern (reason, act, observe), prebuilt vs custom agents, planning and multi-step reasoning, and structured output. Lab: build your first ReAct agent with tools.
-
06 Memory & Persistence (Checkpointers)
Short-term memory through threads, checkpointers (in-memory, SQLite, Postgres), long-term memory stores, and replaying past states. Lab: add memory to your agent.
-
07 Human-in-the-Loop
Interrupts, approval steps, editing state before continuing, and resuming execution — letting a human approve, reject or edit an agent's decision. Lab: build an agent with an approval step.
-
08 Multi-Agent Systems
The supervisor pattern, handoffs between agents, subgraphs, shared vs private state, and choosing an architecture. Lab: build a supervisor team.
-
09 Debugging & Monitoring
Streaming outputs, visualising graphs, LangGraph Studio, tracing with LangSmith, and fixing common problems such as infinite loops and state errors.
-
10 Capstone Project
Design, build and present your own end-to-end agent combining routing, tools, memory and human-in-the-loop, and explain its design decisions.
FAQ
+ Is this course HRD Corp claimable?
Yes, subject to your grant approval. Tarsoft is a registered training provider and will help with the application.
+ Should I take LangChain first?
It isn't required, but The LangChain course gives a good foundation in prompts, tools and RAG before moving on to agents.
+ Can this be run in-house for our team?
Yes — we deliver private cohorts on-site or online, built around your team's use cases.
Related courses
LangChain: Build Smart AI Apps Powered by Your Own Data
Learn LangChain's building blocks — prompts, chains, memory, RAG and agents — and build a chatbot that answers from your own documents.
- 2 days
- Beginner to Intermediate
RM1,500 / pax
Python Intermediate: Secure AI Apps with FastAPI & Ollama
Build a secure RAG assistant that answers questions from your own documents — running 100% on your laptop, with no cloud and no API keys.
- 3 days
- Intermediate
Regular price RM2,000 Early-bird price RM1,800 / pax Early bird
Book 14+ days ahead; RM2,000 regular