Infection is a severe complication of allo-HSCT in children, however, the accurate detection of the infection is hard. In this study, we traced the records of 101 pediatric recipients with allo-HSCT to investigate the pathogens of infection, and collected 54 bronchoalveolar lavage fluid, 32 blood, and 15 cerebrospinal fluid samples. In these samples, 87 was with post-transplant infection and 14 without infection. Using the metagenomic next-generation sequencing (mNGS) and traditional pathogen detection, we compared their sensitivity and specificity to detect pathogens of infection. Our results showed that mNGS was more sensitive (89.7%) than conventional pathogen detection (21.8%), with a difference of 67.9% (P < 0.001), However, mNGS was less specific (78.5%) than traditional methods (92.9%), with a difference of 14.4% (P = 0.596). The sensitivity of mNGS for diagnosing pulmonary infections, bloodstream infections or viremia, and CNS infections post-transplant were 91.7, 85.7, and 90.9%, respectively. In contrast, the sensitivity of conventional testing for diagnosing pulmonary infections, bloodstream infections or viremia, and CNS infections post-transplant were 22.9, 21.4, and 18.2%, respectively. There were significant differences in the sensitivity of mNGS and conventional testing in BALF, blood, and CSF samples, with P values of 0.000, 0.000, and 0.002 respectively. Among the patients with pulmonary infection, 11 pathogens were both identified by mNGS and conventional testing, and 33 by mNGS only. The percentage with the mNGS-positive result was 44/48 (91.7%), including viruses (n = 12), bacteria (n = 17), fungi (n = 9) and mixed infections (n = 6). Among the patients diagnosed with fungal pneumonia (n = 9), the most prevalent pathogenic fungi were Pneumocystis jiroveci (n = 6), which were also detected in 4 patients with mixed infectious pneumonia. In the 28 blood specimens of patients with bloodstream infections or viremia, five patients were positive by both mNGS and conventional testing, 19 were positive by mNGS, and 1 was positive by traditional testing only. The percentage with the mNGS-positive results was 24/28 (85.7%), including viruses (n = 12), bacteria (n = 4), fungi (n = 3), and mixed infections (n = 5). Of the 15 CSF specimens enrolled, 11 patients were eventually diagnosed with CNS infections. Ten pathogens were identified by mNGS in the 11 patients, including viruses (n = 8), bacteria (n = 1), and fungi (n = 1). These results suggest that mNGS is more sensitive than conventional pathogen detection for diagnosing infections post HSCT in children which may help the clinic diagnosis. Pneumocystis jiroveci was the most frequent pathogen of pulmonary infections post-transplant, while viruses were the most common pathogens of CNS infections in allo-HSCT recipients.