Aims. Our goal is to provide a wide ultracool companion frequency for metal-poor halo stars for the first time. Methods. We selected nearby halo stars with spectroscopically determined high proper motions and metallicities. We collected first-epoch deep J-band imaging around these stars. We visually compared these images with optical images from public surveys, and the stars with nearby faint sources beyond the optical survey limit were followed up by deep J-band imaging with baselines of two to four years. We searched for comoving sources between the two epochs. Results. Our observation reached an average depth of Jlim = 22.8 and 23.0 mag (Vega) in the first and the second epoch, respectively, enabling us to detect esds earlier than type esdT0 at a distance up to 250 pc. With this depth, we found no bona fide wide ultracool comoving companions to a nearby metal-poor halo star within a typical projected separation range of a few hundred to a few thousand au from the star. We place an upper limit of the frequency of wide ultracool companions to metal-poor halo stars of 4.0% at a confidence level of 90%. With four stellar comoving companions found and confirmed by Gaia, we derived a wide stellar companion frequency of 6.1+7.2−4.0% 6 . 1 − 4.0 + 7.2 % at a confidence level of 90%. Conclusions. We conclude that wide ultracool companions are rare around metal-poor halo stars, with a frequency marginally lower than that of wide stellar companions. We claim that no metallicity dependence is found for the wide ultracool companion frequency around stars so far. Formation and retention processes in binary systems are likely to operate less efficiently for ultracool secondaries.