This paper addresses robust model predictive control (MPC) for time-delay systems with polytopic uncertainty.Uncertain time-varying input delay and state delays are considered, and the infinite horizon control moves areparametrised into an augmented state feedback law at each time instant. A receding horizon implementation ofthis state feedback law renders satisfaction of input/state constraints and closed-loop stability. For time-invariantdelays and known delays, simplified results are obtained. A numerical example and a benchmark problem oncontinuous stirred tank reactor (CSTR) are given to illustrate the effectiveness of the proposed techniques.