A small class of RNA molecules, in particular the tiny genomes of viroids, are circular. Yet most structure prediction algorithms handle only linear RNAs. The most straightforward approach is to compute circular structures from “internal” and “external” substructures separated by a base pair. This is incompatible, however, with the memory-saving approach of the Vienna RNA Package which builds a linear RNA structure from shorter (internal) structures only. Here we describe how circular secondary structures can be obtained without additional memory requirements as a kind of “post-processing” of the linear structures.