A variety of strategies to incorporate unnatural amino acids into proteins have been pursued, but all have limitations with respect to technical accessibility, scalability, applicability to in vivo studies, or site specificity of amino acid incorporation. The ability to selectively introduce unnatural functional groups into specific sites within proteins, in vivo, provides a potentially powerful approach to the study of protein function and to large-scale production of novel proteins. Here we describe a combined genetic selection and screen that allows the rapid evolution of aminoacyl-tRNA synthetase substrate specificity. Our strategy involves the use of an "orthogonal" aminoacyl-tRNA synthetase and tRNA pair that cannot interact with any of the endogenous synthetase-tRNA pairs in Escherichia coli. A chloramphenicol-resistance (Cm(r)) reporter is used to select highly active synthetase variants, and an amplifiable fluorescence reporter is used together with fluorescence-activated cell sorting (FACS) to screen for variants with the desired change in amino acid specificity. Both reporters are contained within a single genetic construct, eliminating the need for plasmid shuttling and allowing the evolution to be completed in a matter of days. Following evolution, the amplifiable fluorescence reporter allows visual and fluorimetric evaluation of synthetase activity and selectivity. Using this system to explore the evolvability of an amino acid binding pocket of a tyrosyl-tRNA synthetase, we identified three new variants that allow the selective incorporation of amino-, isopropyl-, and allyl-containing tyrosine analogs into a desired protein. The new enzymes can be used to produce milligram-per-liter quantities of unnatural amino acid-containing protein in E. coli.