The traditional picture of evolutionary dynamics is that genomes live in sequence space and adapt on fitness landscapes. Mutation rates are considered to be constant or externally regulated. If, however, we take into account that genomes also encode for enzymes that perform replication and error correction, then individual genomes not only have a specific replication rate (fitness), but also a mutation rate. This leads to the concept of a mutation landscape. We explore evolution on mutation landscapes. Localization in pure mutation landscapes is only possible under extremely restrictive conditions. Coupling of mutation landscapes and fitness landscapes leads to localization and hence adaptation and evolution. We analyze how mutation landscapes facilitate localization in fitness landscapes and vice versa. We also analyze the optimal mutation landscape that maximizes the rate of transitions between isolated fitness peaks.