Genomic selection leverages historical phenotype data from a breeding program to more accurately predict the genetic value of related individuals. In plant breeding, historical data are usually unbalanced with respect to individuals, and the best model to account for micro-environmental variation often varies across trials. These complexities are readily accommodated using a two-stage approach, in which best linear unbiased estimates (BLUEs) of genotypic value within each environment are computed in Stage 1 and then used as the response variable in Stage 2. To fully utilize the data, the variance and covariances of the estimates from Stage 1 must be included in Stage 2, but not many software packages can do this. R package polyBreedR was developed to facilitate genomics-assisted breeding in autopolyploid and diploid species, and the objective of this research was to add a user-friendly interface for two-stage analysis of multi-environment trials. The package uses ASReml-R for variance component estimation and contains vignettes to illustrate several workflows, using data from the University of Wisconsin potato breeding program. In one vignette, trial data from six years at one location are analyzed to predict breeding values and estimate selection accuracy. A second vignette covers the estimation of genetic correlations between traits and locations and the construction of selection indices.