Directed evolution is a powerful technique to tailor a biomolecule for a desired application. Development cycles typically entail the generation of DNA libraries from a starting gene, the expression of different protein variants, and finally the identification of the best variants through screening. A commonly employed method for protein diversity generation is site saturation mutagenesis, the mutation of one or more residues to each of the 20 proteinogenic amino acids. We reimagine this diversity generation process through the lens of an “orthogonal central dogma”, and develop tools for the generation of protein variants by translation-level mutagenesis. We engineer a suite of genetic code expansion tools to efficiently refactor a set of stop and/or sense codons of a genomically recoded organism and demonstrate efficient encoding of the 20 proteinogenic amino acids at a single recoded codon or subsets of amino acids at multiple sites into model proteins. We apply the toolkit to different protein engineering challenges, including the optimization of fluorescent proteins, a self-labeling protein, and an oxidative enzyme. We anticipate that our approach will augment current protein engineering workflows and aid in the exploration of sequence-function relationships of proteins.