Martin,
...What it is for? Is it this one?
"Free & open source, high-performance, distributed memory object caching system, generic in nature, but intended for use in speeding up dynamic web applications by alleviating database load."
-- https://memcached.org/...
A good example is the use-case of caching DB query results, see
https://docs.memcached.org/userguide/More specifically, the API (if I am correct) enables your client side application to query up the results in a memcached server (which could be one, or many of them) to see whether the results for that particular query have been previously stored and have a specific TTL assigned to them.
If the HIT is there, great, you do not end up putting the same query execution load on the DBMS server itself, instead you simply pull that result from it's "temporary" storage on the memcached server.
I would consider this to be an architectural-like tool where you would want to provision it's use right from the very start. I am not quite sure how you'd make use of it in a typical client side user environment though, and I do not think its intended to do that either.
@Igor, how are you using this?