Central nervous system Infections (CNSIs) is a disease characterized by complex pathogens, rapid disease progression, high mortality rate and high disability rate. Here, we evaluated the clinical value of metagenomic next generation sequencing (mNGS) in the diagnosis of central nervous system infections and explored the factors affecting the results of mNGS. We conducted a retrospective study to compare mNGS with conventional methods including culture, smear and etc. 111 suspected CNS infectious patients were enrolled in this study, and clinical data were recorded. Chi-square test were used to evaluate independent binomial variables, taking p < 0.05 as statistically significant threshold. Of the 111 enrolled cases, 57.7% (64/111) were diagnosed with central nervous system infections. From these cases, mNGS identified 39.6% (44/111) true-positive cases, 7.2% (8/111) false-positive case, 35.1% (39/111) true-negative cases, and 18.0% (20/111) false-negative cases. The sensitivity and specificity of mNGS were 68.7% (44/64) and 82.9% (39/47), respectively. Compared with culture, mNGS provided a higher pathogen detection rate in CNSIs patients (68.7% (44/64) vs. 26.5% (17/64), p < 0.0001). Compared to conventional methods, positive percent agreement and negative percent agreement was 84.60% (44/52) and 66.1% (39/59) separately. At a species-specific read number (SSRN) ≥ 2, mNGS performance in the diagnosis of definite viral encephalitis and/or meningitis was optimal (area under the curve [AUC] 0.758, 95% confidence interval [CI] 0.663–0.854). In bacterial CNSIs patients with significant CSF abnormalities (CSF WBC > 300*10 6/L), the positive rate of CSF mNGS is higher. To sum up, conventional microbiologic testing is insufficient to detect all neuroinvasive pathogens, and mNGS exhibited satisfactory diagnostic performance in CNSIs and with an overall detection rate higher than culture ( p < 0.0001).