How to create universes with internal flux

String compactifications typically require fluxes, for example in order to stabilise moduli. Such fluxes, when they thread internal dimensions, are topological in nature and take on quantised values. This poses the puzzle as to how they could arise in the early universe, as they cannot be turned on incrementally. Working with a string inspired model in 8 dimensions, we show that there exist no-boundary solutions in which internal fluxes are present from the creation of the universe onwards. The no-boundary proposal can thus explain the origin of fluxes in a Kaluza-Klein context. In fact, it acts as a selection principle since no-boundary solutions are only found to exist when the fluxes have the right magnitude to lead to an effective potential that is positive and flat enough for accelerated expansion. Within the range of selected fluxes, the no-boundary wave function assigns higher probability to smaller values of flux. Our models illustrate how cosmology can act as a filter on a landscape of possible higher-dimensional solutions. If time permits, I will include comments about an allowability criterion for complex metrics and how it might affect these probabilities.