Abstract Multiplanet systems orbiting M dwarfs provide valuable tests of theories of small-planet formation and evolution. K2-3 is an early M dwarf hosting three small exoplanets (1.5–2.0 R ⊕ ) at distances of 0.07–0.20 au. We measure the high-energy spectrum of K2-3 with HST/COS and XMM-Newton and use empirically driven estimates of Ly α and extreme-ultraviolet flux. We use EXOFASTv2 to jointly fit radial velocity, transit, and spectral energy distribution data. This constrains the K2-3 planet radii to 4% uncertainty and the masses of K2-3b and c to 13% and 30%, respectively; K2-3d is not detected in radial velocity measurements. K2-3b and c are consistent with rocky cores surrounded by solar composition envelopes (mass fractions of 0.36 − 0.11 + 0.14 % and 0.07 − 0.05 + 0.09 % ), H 2 O envelopes ( 55 − 12 + 14 % and 16 − 10 + 17 % ), or a mixture of both. However, based on the high-energy output and estimated age of K2-3, it is unlikely that K2-3b and c retain solar composition atmospheres. We pass the planet parameters and high-energy stellar spectrum to atmospheric models. Dialing the high-energy spectrum up and down by a factor of 10 produces significant changes in trace molecule abundances, but not at a level detectable with transmission spectroscopy. Though the K2-3 planets span the small-planet radius valley, the observed system architecture cannot be readily explained by photoevaporation or core-powered mass loss. We instead propose that (1) the K2-3 planets are all volatile-rich, with K2-3d having a lower density than typical of super-Earths, and/or (2) the K2-3 planet architecture results from stochastic processes such as planet formation, planet migration, and impact erosion.