In the global fight against antibiotic resistance, the need for alternatives is more pressing than ever. Antimicrobial peptides (AMPs), short oligopeptides usually produced as part of the immune system of a host, have shown great promise against resistant bacteria, biofilms and even cancer cells. Engineering AMPs that are both specific to a set of bacteria and stable is among the main challenges of the field. Herein, we propose two separate tools to support these efforts. The first one is an AMP SECretion system based in E. coli, dubbed AMPSEC, that can be used to produce active AMPs with specific targets (i.e., gram-positive bacteria or any specific specie). This recombinant protein system uses surface display technologies coupled with specific protease activity to express, export, and release functional AMPs that could readily affect neighbouring target bacteria. The AMPSEC would be ideal to screen AMP libraries, removing the need for purification or chemical synthesis in order to observe toxicity. It could also be used for AMP production, where the secreted AMPs would be purified from the growth medium by HPLC. Finally, if the recombinant system is inserted in a probiotic host, it might even be useful to deliver AMPs in the gut to treat dysbiosis. Herein, we explored six surface display apparatuses for their applicability for AMPSEC and found three out of the six being fully functional in transporting the cargo although the cleaving activity needs to be coordinated better with its localization at the outer membrane. A robust proof-of-concept workflow has also been established and used to evaluate the performance of those six display apparatuses. The second is a bioinformatics approach to highlighting the relationship between the primary structure and the microbial target specificity of AMPs. Our method first clusters sequences from the DRAMP AMP repository using the Linclust algorithm. De novo motif discovery tools can then extract AMP sequence motifs relating to target specificity. These motifs could guide randomized sequence AMP library creation and decrease the number of inactive sequences generated. Clustering AMPs, however, proved to be rather challenging due to the large sequence length variation in the databases, the small sample size and their overall short lengths. It would then be necessary to design an algorithm suited to handle this specific kind of proteomics dataset. A library eventually created using the discovered motifs could then be used with AMPSEC. Combined, these two tools will further improve our ability to design stable AMPs targeting specific bacteria.