Background: Biomarkers have been widely proposed as having potential to improve tuberculosis (TB) detection, prediction of progression, and treatment monitoring, but evidence is fragmented and it is unclear which biomarkers have the strongest evidence base. We undertook an umbrella review to synthesise this evidence.Methods: This umbrella review (PROSPERO CRD420251230142) evaluated systematic reviews of biomarkers for three broad TB use-cases: detection of TB disease, prediction of progression to TB disease, and TB treatment monitoring and optimisation. We searched MEDLINE, Web of Science, Global Index Medicus and Cochrane Library for systematic reviews of host- or pathogen-derived biomarkers from January 2018 and appraised review quality using AMSTAR-2. For each biomarker class and use case we summarised the technology readiness, strength of evidence, diagnostic accuracy, extent of independent validation, operational characteristics, and unmet need addressed. We synthesised evidence for each biomarker class qualitatively and quantitatively, against a pre-specified biomarker prioritisation framework.Findings: Forty-five systematic reviews met inclusion criteria: 39 on TB detection, 7 on progression and 6 on treatment monitoring. Nine classes of host- or pathogen-based biomarkers were identified, most commonly host protein, host transcriptomic and host miRNA. For TB detection, evidence was strongest for host protein, host transcriptomic, and pathogen antigen classes, while the immune phenotype, exhaled breath and pathogen DNA/RNA classes were earlier in development, lacking independent validation. Across use cases, estimates of diagnostic accuracy varied widely and were often lower in higher quality studies with clinically relevant populations. In high-quality studies no biomarkers consistently met WHO target product profile thresholds. For TB progression, the strongest evidence was for host transcriptomic, which predicted the short-term risk of TB progression but showed modest performance over longer follow-up intervals. For TB treatment monitoring, evidence to predict clinical outcomes is limited across all classes.Interpretation: Many systematic reviews have been published describing numerous TB biomarkers, especially for the TB detection use case. Robust validation of candidate biomarkers in clinically relevant populations is lacking for many classes. Further prospective head-to-head validation studies are needed to determine which biomarkers have meaningful potential to improve individual and public health outcomes.