Anthropic has published a very interesting article about building AI agents. Have you read it? Anthropic engineers suggest that you should understand the underlying code before using any abstraction frameworks like LangGraph; otherwise, it may get hard for you to debug and understand your own code. Also, they suggest that you should only add complexity when it's really necessary for your use case. Simplicity is the best, and you shouldn't build agents when you don't need them. They also provided a detailed overview of commonly used agentic systems. Have a look at it 👉 https://www.anthropic.com/engineering/building-effective-agents