It's actually quite a bit more than REBL; the description downplays something quite significant ... it can run as a client against a running Clojure in a separate process. Although that sounds simple ("it's just data") the number of edge cases involving everything from infinite lists, to dealing with functions as elements in the data, to differences in what's on the classpath between client and server, has impacted the implementation.
Wow, clojure devs are learning that ides are useful