Enzymes empower chemical industries and are the keystone for metabolic engineering. For example, linalool synthases (LSs) are indispensable for the biosynthesis of linalool, an important fragrance used in 60-80% cosmetic and personal care products. However, plant LSs have low activities while expressed in microbes. Aided by bioinformatics analysis, four linalool/nerolidol synthases (LNSs) from various Agaricomycetes were accurately predicted and validated experimentally. Furthermore, we discovered a novel LS with exceptionally high levels of selectivity and activity from Agrocybe pediades, ideal for linalool bioproduction. It effectively converted glucose into enantiopure (R)-linalool in Escherichia coli, 44-fold and 287-fold more efficient than bacterial and plant counterparts, respectively. Phylogenetic analysis indicated the divergent evolution paths for plant, bacterial and fungal LSs. More critically, structural comparison provided catalytic insights in Ap.LS superior specificity and activity, and mutational experiments validated the key residues responsible for the specificity.