Welcome to Building with Agents

If you've already built a few demo agents (or read AI Agents 101), you know they can do impressive things. You also know they break in production. The loop goes off the rails. The tool calls fail silently. The cost spikes. Errors get swallowed. Two minutes into the run, you have no idea what the agent is doing. This course is about closing that gap. Building agents that actually work in production is a real engineering discipline, distinct from prompting and distinct from traditional backend development. The patterns are still being figured out, but a clear toolkit has emerged: LangChain for simple cases, LangGraph for stateful workflows, durable state for resumability, structured observability, evaluation pipelines. We'll cover: • Picking a framework: LangChain, LangGraph, CrewAI, AutoGen • Your first agent with LangChain (developer guide) • Adding RAG: the most common production pattern • Choosing a model: Claude vs GPT-4 for agents • LangGraph: why production teams use it • Hands-on LangGraph in 13 steps • When to use LangChain vs LangGraph (decision guide) • Building agents in TypeScript (full type safety) • The official LangGraph repo (advanced patterns) • Anthropic on building effective agents (the principles from the team behind Claude) This course is for developers and ML engineers. You should be comfortable with Python (or TypeScript) and have built at least a simple LLM-powered feature before. Minimum karma 10: this is intermediate to advanced. Take AI Agents 101 first if you're new. Time: ~4 to 5 hours across 10 articles.