Objective. This work aims at characterizing LiF:Mg,Ti thermoluminescence detectors (TLDs) for dosimetry of a 250 MeV proton beam delivered at ultra-high dose rates (UHDR). Possible dose rate effects in LiF:Mg,Ti, as well as its usability for dosimetry of narrow proton beams are investigated. Approach. LiF:Mg,Ti (TLD-100^TM Microcubes, 1 mm × 1 mm × 1 mm) was packaged in matrices of 5 × 5 detectors. The center of each matrix was irradiated with single-spot low-LET (energy >244 MeV) proton beam in the (1–4500) Gy s^-1 average dose rates range. A beam reconstruction procedure was applied to the detectors irradiated at the highest dose rate (Gaussian beam sigma <2 mm) to correct for volumetric averaging effects. Reference dosimetry was carried out with a diamond detector and radiochromic films. The delivered number of protons was measured by a Faraday cup, which was employed to normalize the detector responses. Main results. The lateral beam spread obtained from the beam reconstruction agreed with the one derived from the radiochromic film measurements. No dose rates effects were observed in LiF:Mg,Ti for the investigated dose rates within 3% (k = 1). On average, the dose response of the TLDs agreed with the reference detectors within their uncertainties. The largest deviation (-5%) was measured at 4500 Gy s⁻¹. Significance. The dose rate independence of LiF:Mg,Ti TLDs makes them suitable for dosimetry of UHDR proton beams. Additionally, the combination of a matrix of TLDs and the beam reconstruction can be applied to determine the beam profile of narrow proton beams.