Abstract Under the framework of network artificial chemistry (NAC), a method to construct a phenotypic machine from genotypic information is proposed. The genotype is expressed as a sequence of nodes (a node chain) with symbol data. According to an implemented algorithm, a chain is agglomerated, tangled, and finally folded into a node cluster that works as a control-flow machine in the network. Two examples of control-flow clusters, splitase and replicase, are presented and their functionality is demonstrated. Several meanings and future research directions are discussed.