Just throwing my 2c onto the table. :)
Since the CFEngine server is not much more than a file server, it's incredibly lightweight, so running without one doesn't seem to provide a lot of benefit, especially since it means some other mechanism has to setup on all clients in its absence.
If you're using the community edition, something as lightweight as a Raspberry Pi can easily support a "personal configuration environment". It's still a good idea to have the masterfiles be in a git repository, but if you have a server instance, then you only have to handle the updates on a single system, via a cron job or just a manual git pull if you're not updating very often.
To Aleksey's point, if you're in a "home lab" style environment, definitely consider using the Enterprise version. I have a smallish VMware cluster of Intel NUCs, and run Enterprise as a VM for the rest of the environment. Also, just because you're running Enterprise doesn't mean you can't use community clients. In my case, I have a small pile of Raspberry Pi systems, and since there isn't an Enterprise client for ARM ( by the way, if an Enterprise ARM build existed, that would really cool ), I just run the community version on them, point them at the Enterprise VM for their policy server, and all is well.
And in the case where I'm just running CFEngine on a single machine, I _still_ prefer to run a server. For example, a couple jobs ago I used CFEngine for an HPC cluster. To work out new policy, I would fire up a VM, install CFEngine, clone the masterfiles, bootstrap the system to itself, and roll from there. If I wanted to have it be a little more of a test environment, it was just a matter of firing up one or more VMs, and pointing them at the already running VM for their policy server.
;tldr The policy server piece of CFEngine is extremely lightweight. It's nowhere close to the headache involved in standing up a Puppet Master for a home lab style environment. :)
Cheers,
sean