The aim of this paper is to solve the inconsistency problem à la Barro and Gordon within a New Keynesian model and to derive time-consistent (stable) interest rate rules of Taylor-type. We find a multiplicity of stable rules. In contrast to the Kydland/Prescott-Barro/Gordon approach, implementing a monetary rule where the cost and benefit resulting from inconsistent policy coincide - which implies a net gain of inconsistent policy behavior equal to zero - is not optimal. Instead, the solution can be improved by moving into the time-consistent area where the net gain of inconsistent policy is negative. We moreover show that under a standard calibration, the standard Taylor rule is stable in the case of a cost-push shock as well as under simultaneous supply and demand shocks.