Skip to main content

LangChain & LangGraph

7 articles · written by Niraj Kumar

LangChain gets you to a working prototype fast and then asks harder questions: where does state live, what happens when a tool call fails halfway through, how do you put a human in the loop without blocking the process, and how do you debug a chain that produced the wrong answer three steps ago.

These posts focus on the answers. They cover LangGraph's graph model for multi-step agents, PostgreSQL checkpointers for durable state, human-in-the-loop approval gates, tool design patterns that keep agents from looping, and the structural choices that separate a demo agent from one you can leave running.

There is also a comparison with LlamaIndex for readers still choosing a framework, and several posts that use LangChain purely as the retrieval layer of a larger RAG system rather than as the whole application.

Need this built rather than explained?

Production LangChain & LangGraph systems for RAG, agents, and AI workflows.

LangChain Developer Services

Related topics

← All articles