In statistics, how is regression able to be used for causative inferences, yet linear correlation is not?

I understand that correlation and causation are not the same, however, what does regression do differently that allows us to make that leap? Especially considering that it does not matter which variable we use as the y and the x.