The cyanelles of the glaucocystophyte alga Cyanophora paradoxa resemble endosymbiotic cyanobacteria, especially in the presence of a peptidoglycan wall between the inner and outer envelope membranes. However, it is now clear that cyanelles are in fact primitive plastids. Phylogenetic analyses of plastid, nuclear and mitochondrial genes support a single primary endosymbiotic event. In this scenario, cyanelles and all other plastid types are derived from an ancestral photosynthetic organelle combining the high gene content of rhodoplasts and the peptidoglycan wall of cyanelles. This means that the import apparatuses of all primary plastids, i.e. those from glaucocystophytes, red algae, green algae and higher plants, should be homologous. If this is the case, then transit sequences should be similar and heterologous import experiments feasible. Thus far, heterologous in vitro import has been shown in one direction only: precursors from C. paradoxa were imported into isolated pea or spinach chloroplasts. Cyanelle transit sequences differ from chloroplast stroma targeting peptides in containing in their N-terminal domain an invariant phenylalanine residue which is shown here to be crucial for import. In addition, we now demonstrate that heterologous precursors are readily imported into isolated cyanelles, provided that the essential phenylalanine residue is engineered into the N-terminal part of chloroplast transit peptides. The cyanelle and likely also the rhodoplast import apparatus can be envisaged as prototypes with a single receptor/channel showing this requirement for N-terminal phenylalanine. In chloroplasts, multiple receptors with overlapping and less stringent specificities have evolved, explaining the efficient heterologous import of native precursors from C. paradoxa.