Metagenomics has the potential to revolutionise infectious diseases diagnostics, from rapid species and antimicrobial resistance prediction, to finding unrecognised and sometimes untreated infections. Our aim was to summarise all literature on culture-independent metagenomic sequencing to describe the accuracy of species and antimicrobial resistance prediction and, describe the challenges and progress in the field. We conducted a systematic review with meta-analysis from eligible studies retrieved from PubMed, Google Scholar and bioRxiv and, assessed risk of bias and quality using the QUADAS-2 tool. This study is registered with PROSPERO, number CRD42020163777. We identified 36 studies, 22 of which used a species-agnostic approach to identify all possible pathogens. In these studies, the overall sensitivity and specificity of pathogen species detection were 88% (95%CI 81-92%) and 86% (95%CI 70-94%) respectively. Antimicrobial resistance prediction and comparison to phenotypic results was undertaken in six studies. Categorical agreement was 83% (95%CI 68-92%), very major (prediction sensitive, phenotype resistant) and major error (prediction resistant, phenotype sensitive) rates were 9% (95%CI 2-27%) and 1% (95%CI 0-20%) respectively. We report limited use of negative controls in studies 61% (22/36) which contribute to a major challenge of discriminating true pathogens from contamination, where there is no convergence on methodology. More efficient human DNA depletion methods are required as a median of 79% (IQR 62-96) [Range 7-98] of sequences were classified as human despite laboratory depletion techniques. The median time from sample to result was 23.5 hours (7-31) [4-144], with sequencing time accounting 10 hours (4.8-16) [1-16]. The average reported consumables cost per sample ranged from $128 to $685. The science and regulatory environment are rapidly developing, and its role as a routine test or test of last resort still needs to be determined, however it is likely that clinical metagenomics will be an increasing part of the clinician′s armamentarium to diagnose infectious diseases in the near future.