Agents

Building intelligent, stateful AI agents with LangGraph. This section covers the agent lifecycle — from foundations and tool calling through multi-agent orchestration, memory, context engineering, and production harnesses.

Topics

• LangGraph Foundations — Core concepts: graphs, nodes, edges, and message-centric state management for stateful AI workflows

• Tool Calling — How LLMs invoke external tools via function calling, structured outputs, and the Model Context Protocol

• Agentic Patterns — ReAct, reflection, planning, multi-expert orchestration, and agentic RAG patterns (CRAG, Self-RAG, Adaptive RAG)

• Multi-Agent Collaboration — Designing systems where multiple specialized agents collaborate through orchestration, handoff, and context injection

• Human-in-the-Loop — Integrating human approval steps into agent workflows with breakpoints, checkpointing, and state persistence

• Model Context Protocol (MCP) — The open standard for connecting LLMs to tools and data sources, adopted by all major AI providers

• Agent Memory Systems — Memory architectures for persistent, context-aware agents: working, semantic, episodic, and procedural memory

• Context Engineering — Designing and managing context windows for optimal LLM performance, including prompt caching

• Harness Engineering — Shaping the operational environment around agents for reliable, safe, and observable autonomous work

Prerequisites

Complete Foundations and RAG Optimization first.

• LangGraph Foundations & State Management
• Tool Calling & Tavily Search
• Multi-Expert Research Agent with ReAct Pattern
• Multi-Agent Collaboration
• Human-in-the-Loop & Persistence
• Model Context Protocol (MCP) NEW
• Agent Memory Systems NEW
• Context Engineering
• Harness Engineering
• Practice — Agents
• Assessment — Agents