Thousands of pathogens are known to infect humans, but only a fraction are readily identifiable using current diagnostic methods. Microbial cell-free DNA sequencing offers the potential to non-invasively identify a wide range of infections throughout the body, but challenges to clinical-grade metagenomic testing must be addressed. Here we describe the analytical and clinical validation of a next generation sequencing test that identifies and quantifies microbial cfDNA in plasma from 1,250 clinically relevant bacteria, DNA viruses, fungi, and eukaryotic parasites. Test accuracy, precision, bias and robustness to a number of metagenomics-related challenges were assessed using panel of 13 microbes that model key determinants of performance in 358 contrived plasma samples, along with 2,625 in silico simulated datasets and 580 clinical study samples. The test showed 93.7% positive agreement with blood culture in a cohort of 350 patients suspected of sepsis, and identified an independently adjudicated cause of the sepsis alert more often than all microbiology testing combined (169 etiological determinations versus 132). Analysis of the first 1,500 patient samples tested in the CLIA lab showed that more than 80% of results were delivered the day after sample receipt, with 52.5% of reports identifying at least one microorganism.