Most graph databases I'm aware of require a server and aren't embedded into applications. Using datalog and applying it to graphs also more general and thus preferable than cypher which is still a niche specifically for graph databases.

Also "because they can" seems obligatory.