Diabetes mellitus is a global epidemic with major impacts on human health and society. Drug discovery for diabetes can be facilitated by the development of a rapid, vertebrate-based screen for identifying new insulin mimetic compounds. Our study describes the first development of a zebrafish-based system based on direct monitoring of glucose flux and validated for identifying novel anti-diabetic drugs. Our system utilizes a fluorescent-tagged glucose probe in an experimentally convenient 96-well plate format. To validate our new system, we identified compounds that can induce glucose uptake via activity-guided fractionation of the inner shell from the Japanese Chestnut (Castanea crenata). The best performing compound, UP3.2, was identified as fraxidin and validated as a novel insulin mimetic using a mammalian adipocyte system. Additional screening using sets of saponin- and triazine-based compounds was undertaken to further validate this assay, which led to the discovery of triazine PP-II-A03 as a novel insulin mimetic. Moreover, we demonstrate that our zebrafish-based system allows concomitant toxicological analysis of anti-diabetic drug candidates. Thus, we have developed a rapid and inexpensive vertebrate model that can enhance diabetes drug discovery by preselecting hits from chemical library screens, before testing in relatively expensive rodent assays.