Through VMware Tools driver, both nworks MP and SPI can monitor logical partitions free space (c:\ d:\ etc) and check presence of GuestOS heartbeat. This is quite basic monitoring of course.
If there is a need to monitor eventlog, do services probing, collect metrics and so on, then an agent is of course the only option.
You can reduce the number of systems with agents installed on them for RMS performance optimisation first thing, not speaking about licensing. If there's only "Core OS" MP installed in significant agent rollout, and your monitoring focused on performance first priority, metrics collected from within the guest would anyway be incorrect, as agent wouldn't be showing CPU Ready time, Memory Balloon driver inflation. This is what we call a "tunnel vision".