Let's suppose that your server S exposes an API and that this API is used by many clients, or maybe just one — it doesn't really matter.
Let's focus on a client only, let's call it C. Documents were exchanged between S and C, so to agree on the interface.
After that, tests were done and none raised issues or, if they did, they were fixed.
The tests stated that S and C can talk, that they can understand each other, that C calls S correctly, that S acts properly according to the request, that C parses server's answer correctly, and so on.
Everything works fine, or mostly fine, since the beginning.