I tried it for about a week last year, back when I was going through a bunch of FOSS engines/frameworks to work out what I wanted to stick with long-term. (Warning: Link contains strong opinions that may differ in some places from most devs I know)
To save you the trouble, I can give a basic overview of how it felt to use and why I chose not to go with it.
In terms of how the API feels, I'd say it's got the same kind of feel as love2d (which I currently use). It's a relatively low-level API that handles much of the boilerplate and gives you easy access to asset loading / audio / rendering / etc, but doesn't make many assumptions about code structure. Which, in my opinion, is perfect for a framework!
As far as problems go, there's two main ones that I ran into. First of all, it's designed for C and if you use it in C you will have to deal with C-based issues. For instance, asset loading code takes in a filepath but doesn't give you any platform-independent solutions for actually locating files on disk (remember, relative paths are not relative to the exe but to the working dir). You'll need to provide stuff like that, data structures, etc most likely if you're not working in a higher-level language.
The other main issue I found was that there's no 'proper' API docs. They've got example code and a cheatsheet, but the latter doesn't cover the full API so I wound up having to read header files to find things which... kinda sucks? Bad documentation can be a serious time sink, especially when starting out.
So yeah, honestly it's gonna come down to what you want to make and what you value as a developer. I can see Raylib working for some folks, but in my personal case I found the cons outweighed the pros and moved on.