BACKGROUND: Appropriate-technology tests are needed for Mycobacterium tuberculosis drug-susceptibility testing (DST) in resource-constrained settings. We evaluated the MDR/XDRTB colour plate thin-layer agar test (TB-CX) for M. tuberculosis DST by directly testing sputum at University of Gondar Hospital. METHODS: Sputum samples were each divided into 2 aliquots. One aliquot was mixed with disinfectant and applied directly to the TB-CX quadrant petri-plate containing culture medium with and without isoniazid, rifampicin or ciprofloxacin. Concurrently, the other aliquot was decontaminated with sodium hydroxide, centrifuged and cultured on Lowenstein-Jensen media, then the stored M. tuberculosis isolates were sub-cultured in BACTEC Mycobacteria Growth Indicator Tube (MGIT) 960 for reference DST. RESULTS: TB-CX text yielded DST results for 94% (123/131) of positive samples. For paired DST results, median days from sputum processing to DST was 12 for TB-CX versus 35 for LJ-MGIT (P < 0.001). Compared with LJ-MGIT for isoniazid, rifampicin and MDR-TB, TB-CX test had: 59%, 96% and 95% sensitivity; 96%, 94% and 98% specificity; and 85%, 94% and 98% agreement, respectively. All ciprofloxacin DST results were susceptible by both methods. CONCLUSION: The TB-CX was simple and rapid for M. tuberculosis DST. Discordant DST results may have resulted from sub-optimal storage and different isoniazid concentrations used in TB-CX versus the reference standard test.