Drug resistance in Mycobacterium tuberculosis is relentlessly progressing while only a handful of novel drug candidates are developed. Here we describe a GFP-based high-throughput screening of 386,496 diverse compounds to identify putative tuberculosis drug candidates. In an exploratory analysis of the model organism M. bovis BCG and M. smegmatis and the subsequent screening of the main library, we identified 6354 compounds with anti-mycobacterial activity. These hit compounds were predominantly selective for mycobacteria while dozens had activity in the low μM range. We tested toxicity against the human monocyte/macrophage cell line THP-1 and elaborated activity against M. tuberculosis growing in liquid broth, under unique conditions such as non-replicating persistence or inhibition of M. tuberculosis residing in macrophages. Finally, spontaneous compound-resistant M. tuberculosis mutants were selected and subsequently analyzed by whole genome sequencing. In addition to compounds targeting the well-described proteins Pks13 and MmpL3, we identified two novel scaffolds targeting the bifunctional guanosine pentaphosphate synthetase/ polyribonucleotide nucleotidyltransferase GpsI, or interacting with the aminopeptidase PepB, a probable pro-drug activator. KEY MESSAGES: A newly identified scaffold targets the bifunctional enzyme GpsI. The aminopeptidase PepB is interacting with a second novel scaffold. Phenotypic screenings regularly identify novel compounds targeting Pks13 and MmpL3.