We develop the theory of SDE driven by nonlinear Lévy noise, aiming at applications to Markov processes. It is shown that a conditionally positive integro-differential operator (of the Lévy-Khintchine type) with variable coefficients (diffusion, drift and Lévy measure) depending Lipschitz continuously on its parameters generates a Markov semigroup, where the measures are metricized by the Wasserstein-Kantorovich metrics W_p . The analysis of SDE driven by nonlinear Lévy noise was initiated by the author in (Kolokoltsov, Probability Theory Related Fields, DOI: 10.1007/s00440-010-0293-8, 2009) (inspired partially by Carmona and Nualart, Nonlinear Stochastic Integrators, Equations and Flows, Stochatic Monographs, v. 6, Gordon and Breach, 1990), see also (Kolokoltsov, Nonlinear Markov Processes and Kinetic Equations, Monograph. To appear in CUP, 2010). Here, we suggest an alternative (seemingly more straightforward) approach based on the path-wise interpretation of these integrals as nonhomogeneous Lévy processes. Moreover, we are working with more general W_p -distances rather than with W₂.