Executive function (EF) is a collection of self-regulatory control processes that are compromised by poverty and powerfully predict academic outcomes in children. Despite this, there are few evidence-based interventions to improve EF. Given the importance of measurement of EF in the context of the classroom where children learn, we first report results showing the validity and reliability of over 60,000 web-based, classroom administrations of tests of EF that have previously only been widely used in laboratory research. Using these tests, we next show that 800 min of computer-presented cognitive training exercises can improve EF, after controlling for practice effects and developmental effects (working memory: partial η2 = .039, response inhibition: partial η2 = .132, interference control: partial η2 = .072). The abilities to measure and improve EF at low cost and large scale in classrooms can contribute to improved, evidence-based education and potentially help reduce achievement gaps associated with poverty.