High‐frequency (HF) radars can provide high‐resolution and frequent ocean surface currents observations during tropical cyclone (TC) landfall. We describe the first assimilation of such potential observations using idealized twin experiments with and without these observations. The data assimilation system consists of the Ensemble Adjustment Kalman Filter and a coupled ocean‐atmosphere model. In this system, synthetic HF radar‐observed coastal currents are assimilated, and the 24‐, 48‐ and 72‐hr forecast performances are examined for TCs with various intensities, sizes, and translation speeds. Assimilating coastal surface currents improves the intensity forecast. The errors of the maximum wind speed reduce by 2.7 (33%) and 1.9 m/s (60%) in the 72‐hr forecast and 2.8 (40%) and 1.4 m/s (62%) in the 48‐hr forecast, for Category 4 and 2 cyclones, respectively. These improvements are similar to the current operational TC forecast errors, so that assimilating HF radar observations could be a substantial benefit.