In this paper we propose an iterative algorithm for the solution of Volterra equations of the first kind whose kernel is a square matrix. The algorithm, essentially the Lavrientev method coupled with discretization, is "direct" in the sense that preliminary numerical computation of the derivative of the observed variable is not required. We assume boundedness of the input u and mild regularity conditions of the kernel. We prove convergence of the algorithm in L p (0, T ), 1 ≤ p < +∞ and uniform convergence on the intervals where the input is continuous. Under additional information on u we give both integral and pointwise convergence estimates. The observed variable is read with errors, at discrete time instants.