Chains refer to sequences of calls - whether to an LLM, a tool, or a data preprocessing step. The primary supported way to do this is with LCEL.