LangGraph Agents & Multi-Agent Systems

Build reliable agents as graphs: state, nodes, edges, loops, tools, memory, human approval, streaming, sub-graphs and multi-agent supervisors, with every example run for real offline.

Start course →

Syllabus

Why Agents Need Graphs

  1. Chains, Agents and Workflows
  2. The LangGraph Model: State, Nodes, Edges
  3. Your First Graph
  4. State Reducers: Replace or Accumulate

Control Flow: Branches, Loops, Commands, Fan-Out

  1. Conditional Edges and Routing
  2. Loops: Retry, Refine and Reflect
  3. Recursion Limit: The Safety Net
  4. Command: Update State and Route Together
  5. Fan-Out With Send (Map-Reduce)

Agents With Tools

  1. The Tool-Calling Agent as a Graph
  2. Prebuilt Agents With a Real Model
  3. Designing Tools That Agents Use Well
  4. Agent Limits, Costs and Stop Conditions

Memory, Persistence and Time Travel

  1. Checkpointers and Threads
  2. State History, Replay and Time Travel
  3. Long-Term Memory With a Store
  4. Managing Long Histories: Trim and Summarise

Human Approval, Reliability and Streaming

  1. Human-in-the-Loop With interrupt()
  2. Retry Policies and Error Handling
  3. Streaming Updates, Values and Tokens

Multi-Agent Systems

  1. Why and When to Use Multiple Agents
  2. The Supervisor Pattern
  3. Hand-Offs and Swarm-Style Networks
  4. Sub-Graphs and Hierarchical Teams

Testing, Evaluating and Deploying

  1. Testing Graphs Without a Model
  2. Evaluating Agents: Trajectories, Tools and Outcomes
  3. Deploying, Observability and Guardrails

Putting It Together

  1. Case Study: A Customer-Support Agent Team
  2. Revision: Cheat Sheet and Self-Check