There are some architectures that do kind of a auto configuration.
This may base on broadcasting a request in the LAN and all clients are advised to respond.
This becomes far more tricky if you have dynamically coming and going server/client nodes and you want this to work reliably.
E.g. check the DLNA protocols, UPnP, or on a completely different technology: the FireWire Configuration Mechanisms.
If you go that path by inventing the weel yourself, and if you aim to a reliable system, this approach opens the box of pandora:
- handle slow responding clients
- deal with network dropouts (e.g. WIFI issues)
- deal with re-connecting clients after a client crash or network dropout (the client did not say bye-bye before leaving the network, so, you need to be able to distinguish conflicts and re-connections (use the "bye-bye hello" pattern?)
- nasty timeout issues (don't slow down the configuration by non-cooperative clients)
- avoid broadcast event storms (do not re-configure too often)
So, if you have no real requirement that the server knows up-font all (potential) clients: don't go that path.
It will cost you a lot of time and money to get the sytem running in a decent way.