PolyCLEAN provides a fast and scalable algorithm to perform sparsity-inducing Bayesian imaging, improving upon the proximal solvers in terms of computation time and memory requirements. It applies the Frank-Wolfe algorithm to image reconstruction in radio interferometry. The code achieves a reconstruction quality comparable to modern implementations of CLEAN, with better reconstruction of point sources and diffuse emissions. The code's design takes advantage of sparsity properties of the imaging inverse problem involved in radio interferometry; in particular, when super-resolution was involved, the increased sparsity in the solutions allows PolyCLEAN to outperform optimized CLEAN solvers.