The influence functional method of Feynman and Vernon is used to obtain a quantum master equation for a Brownian system subjected to a Lévy stable random force. The corresponding classical transport equations for the Wigner function are then derived, both in the limit of weak and strong friction. In the theory of Brownian motion one is interested in the time evolution of a system coupled to a large environment. The effect of the coupling is modeled by a stochastic force F (t) with a given probability density P [F (t)]. The dynamics of a Brownian particle of mass M in the presence of an external potential U (x) is then described by the Langevin equationwhere F (t) has been divided into a mean force proportional to the velocity, the friction force −γẋ(t), plus a fluctuating part ξ(t). In the usual treatment of Brownian motion [1], it is assumed that the random force is Gaussian distributed with variance ξ(t)ξ(t ′ ) = 2D δ(t − t ′ ) where D = γkT is the diffusion coefficient, and the Langevin equation is shown to be fully equivalent to a phase-space equation -the Klein-Kramers equation. In the limit of strong friction, the inertial term in the Langevin equation can be neglected and the Klein-Kramers equation reduces to the Smoluchowski equation. However, it has become clear in recent years that many processes in nature, such as anomalous diffusion (for a review see [2-4]), cannot be described by ordinary (Gaussian) Brownian motion. A case in point is the so-called Lévy flight with a stochastic force distributed according to Lévy stable statistics and which has been introduced in connection with super-diffusion [5,6]. In this Letter we consider the generalization of transport equations to describe Lévy Brownian motion. This question has already been addressed in the past by using various methods [5][6][7][8], in particular the CTRW formalism [9,10]. However, all these approaches were limited to coordinate space only. Here we present the first derivation of the Klein-Kramers equation for a Lévy stable process. We consider both the case of a symmetric and asymmetric probability distribution. As our main tool, we employ the influence functional formalism developed by Feynman and Vernon [11,12]. If initially system and environment are not correlated then, according to Feynman and Vernon [11,12], the density operator of the system at time t can be written in coordinate representation asHere the entire information on the coupling to the environment is contained in the influence functionalis the characteristic functional of the probability density P[F(t)], Φ[k(t)] = exp i F (t)k(t)dt P [F (t)]DF (t) .If F (t) is Gaussian distributed with mean F (t) = γ(t) and variance [F (t) − γ(t)][F (t ′ ) − γ(t ′ )] = D(t − t ′ ), the characteristic functional is given by