Context. Strong gravitationally lensed quasars are among the most interesting and useful observable extragalactic phenomena. Because their study constitutes a unique tool in various fields of astronomy, they are highly sought, not without difficulty. Indeed, even in this era of all-sky surveys, their recognition remains a great challenge, with barely a few hundred currently known systems. Aims. In this work we aim to detect new strongly lensed quasar candidates in the recently published Gaia Data Release 2 (DR2), which is the highest spatial resolution astrometric and photometric all-sky survey, attaining effective resolutions from 0.4" to 2.2". Methods. We cross-matched a merged list of quasars and candidates with the Gaia DR2 and found 1,839,143 counterparts within 0.5". We then searched matches with more than two Gaia DR2 counterparts within 6". We further narrowed the resulting list using astrometry and photometry compatibility criteria between the Gaia DR2 counterparts. A supervised machine learning method, Extremely Randomized Trees, is finally adopted to assign to each remaining system a probability of being lensed. Results. We report the discovery of three quadruply-imaged quasar candidates that are fully detected in Gaia DR2. These are the most promising new quasar lens candidates from Gaia DR2 and a simple singular isothermal ellipsoid lens model is able to reproduce their image positions to within $\sim$1 mas. This letter demonstrates the gravitational lens discovery potential of Gaia.