Correct, and even if BOINC is (rather) easy to set up and maintain, maybe (more evaluation is required) we don't even need to bother with this kind of generic infrastructure.
Using programs such as PRPNet/ECMNet as a starting point might do the job, given that those are already dealing with integers, they have a networking protocol between sockets on the two sides, and they launch external worker programs on the client side
Adaptations will be necessary, of course, because we're not dealing with ECM curves or primality tests. And the latest version of ECMNet might be a better starting point than the latest version of PRPNet, because there's not yet any dependency on a SQL server (just edit text files with a simple syntax on the client and the server - I've done this earlier today for local ECMNet testing on an integer).