Direct pathogen detection in blood to diagnose active tuberculosis (TB) has been difficult due to low levels of circulating antigens or due to the lack of specific, high-affinity binding reagents and reliable assays with adequate sensitivity. We sought to determine whether slow off-rate modified aptamer (SOMAmer) reagents with subnanomolar affinity for Mycobacterium tuberculosis proteins (antigens 85A, 85B, 85C, GroES, GroEL2, DnaK, CFP10, KAD, CFP2, RplL, and Tpx) could be useful to diagnose tuberculosis. When incorporated into the multiplexed, array-based proteomic SOMAscan assay, limits of detection reached the subpicomolar range in 40% serum. Binding to native M. tuberculosis proteins was confirmed by using M. tuberculosis culture filtrate proteins and fractions from infected macrophages and via affinity capture assays and subsequent mass spectrometry. Comparison of serum from culture-positive pulmonary TB patients and TB suspects systematically ruled out for TB revealed small but statistically significant (P < 0.0001) differences in the median M. tuberculosis signals and in specific pathogen markers, such as antigen 85B. Samples where many M. tuberculosis aptamers produced high signals were rare exceptions. In concentrated, protein-normalized urine from TB patients and non-TB controls, the CFP10 (EsxB) SOMAmer yielded the most significant differential signals (P < 0.0276), particularly in TB patients with HIV coinfection. In conclusion, direct M. tuberculosis antigen detection proved difficult even with a sensitive method such as SOMAscan, likely due to their very low, subpicomolar abundance. The observed differences between cases and controls had limited diagnostic utility in serum and urine, but further evaluation of M. tuberculosis SOMAmers using other platforms and sample types is warranted.