Granulysin is an antimicrobial peptide (AMP) expressed by human T-lymphocytes and natural killer cells. Despite a remarkably broad antimicrobial spectrum, its implementation into clinical practice has been hampered by its large size and off-target effects. To circumvent these limitations, we synthesized a 29 amino acid fragment within the putative cytolytic site of Granulysin (termed “Gran1”). We evaluated the antimicrobial activity of Gran1 against the major human pathogen Mycobacterium tuberculosis (Mtb) and a panel of clinically relevant non-tuberculous mycobacteria which are notoriously difficult to treat. Gran1 efficiently inhibited the mycobacterial proliferation in the low micro molar range. Super-resolution fluorescence microscopy and scanning electron microscopy indicated that Gran1 interacts with the surface of Mtb, causing lethal distortions of the cell wall. Importantly, Gran1 showed no off-target effects (cytokine release, chemotaxis, cell death) in primary human cells or zebrafish embryos (cytotoxicity, developmental toxicity, neurotoxicity, cardiotoxicity). Gran1 was selectively internalized by macrophages, the major host cell of Mtb, and restricted the proliferation of the pathogen. Our results demonstrate that the hypothesis-driven design of AMPs is a powerful approach for the identification of small bioactive compounds with specific antimicrobial activity. Gran1 is a promising component for the design of AMP-containing nanoparticles with selective activity and favorable pharmacokinetics to be pushed forward into experimental in vivo models of infectious diseases, most notably tuberculosis.