You can either create both VMs as part of a single Vagrantfile and I think this allows them to access each other via hostname on the same "network", or you can assign static IP addresses in the vagrantfile and simply use that address to access each VM. The one thing you'll want to do is make sure if you need passwordless access for provisioning, you can add the vagrant key to each VM (they should already trust it, but they'll need an actual copy to use it to access another machine) and then make your connections using that key. You could also try the ssh-agent-forwarding, but I haven't messed with multi-VM setups very much.