LangGraph

LangGraph models an agent workflow explicitly as a graph — nodes are steps (call a tool, ask the model, check a condition), edges define what happens next, including loops back to an earlier step when a task needs another pass. That explicit structure makes complex, multi-step reasoning easier to debug than an open-ended agent loop, since you can see and control exactly which path execution took.

Built by the LangChain team as the answer to LangChain’s own agent abstractions feeling too implicit for genuinely complex workflows — the two are commonly used together rather than as alternatives.