The Frontier Fields project is an observational campaign targeting six galaxy clusters, with the intention of using the magnification provided by gravitational lensing to study galaxies that are extremely faint or distant. We used the Karl G. Jansky Very Large Array (VLA) at 3 and 6 GHz to observe three Frontier Fields: MACSJ0416.1$-$2403 ($z$ = 0.396), MACSJ0717.5+3745 ($z$ = 0.545), and MACSJ1149.5+2223 ($z$ = 0.543). The images reach noise levels of $sim$1 $mu$Jy beam$^{-1}$ with sub-arcsecond resolution ($sim$2.5 kpc at $z$ = 3), providing a high-resolution view of high-$z$ star-forming galaxies that is unbiased by dust obscuration. We generate dual-frequency continuum images at two different resolutions per band, per cluster, and derive catalogs totalling 1966 compact radio sources. Components within the areas of Hubble Space Telescope and Subaru observations are cross-matched, providing host galaxy identifications for 1296 of them. We detect 13 moderately-lensed (2.1 $<$ $mu$ $<$ 6.5) sources, one of which has a demagnified peak brightness of 0.9 $mu$Jy beam$^{-1}$, making it a candidate for the faintest radio source ever detected. There are 66 radio sources exhibiting complex morphologies, and 58 of these have host galaxy identifications. We reveal that MACSJ1149.5+2223 is not a cluster with a double relic, as the western candidate relic is resolved as a double-lobed radio galaxy associated with a foreground elliptical at $z$ = 0.24. The VLA Frontier Fields project is a public legacy survey. The image and catalog products from this work are freely available.