The detection of small molecules in living cells using genetically encoded FRET sensors has revolutionized our understanding of signaling pathways at the sub-cellular level. However, engineering fluorescent proteins and specific binding domains to create new sensors remains challenging because of the difficulties associated with the large size of the polypeptides involved, and their intrinsically huge conformational variability. Indeed, FRET sensors' design still relies on vague structural notions, and trial and error combinations of linkers and protein modules. We recently designed a FRET sensor for the second messenger cAMP named CUTie (yclic nucleotide Universal ag for maging xperiments), which granted sub-micrometer resolution in living cells. Here we apply a combination of sequence/structure analysis to produce a new-generation FRET sensor for the second messenger cGMP based on Protein kinase G I (PKGI), which we named CUTie2. Coarse-grained molecular dynamics simulations achieved an exhaustive sampling of the relevant spatio-temporal coordinates providing a quasi-quantitative prediction of the FRET efficiency, as confirmed by experiments. Moreover, biochemical characterization showed that the cGMP binding module maintains virtually the same affinity and selectivity for its ligand thant the full-length protein. The computational approach proposed here is easily generalizable to other allosteric protein modules, providing a cost effective-strategy for the custom design of FRET sensors.