LangChain Agents act as controllers capable of using custom tools to expand the capabilities of LLMs. An example would be using customer maths tools (either an Expert Model or own applications) to handle maths problems that LLMs are often bad at. LangChain uses the ReAct framework to power its agents.