It really depends on the API. For stuff that is likely to change it makes sense to abstract away the API calls. eg either because the API provider likes to break stuff or because you need your code to be agnostic towards that particular library eg for portability reasons. But that isn't always a concern and if you're abstracting away an API just for the sake of it then you'll have to question whether you're adding unnecessary complexity and overhead just for the sake of perceived modularity.

Like everything in software development, it's about evaluating the right methodologies for the problems rather than hitting every screw with the same hammer.