Every API has its quirks, sometimes returns different data, has edge cases for certain calls, has different error reporting patterns, rate limiting, authentication methods, sometimes has intended or unintended breaking changes, ....
Sure, you can usually just manually whip up the code for an endpoint or two in a few minutes , but then you need to continually add endpoins, fix the edge cases, adapt to upstream changes, ...
Good bindings make your life a whole lot easier.
But API wrappers handling things like retries, pagination, client-side business logic, and coercion from JSON. Having to account for all of that for two API calls is somewhat tedious and repetitive, and I'm sure we've all solved those problems at least a dozen times.