Metabolism is maintained by complex systems far from equilibrium, which consist of a multitude networks of enzymatic reaction. While the working principles of biology’s regulatory networks are well known, assembling unbalanced enzymatic reaction networks in vitro has proven challenging. The development of synthetic systems with autonomous behavior is therefore limited. Herein, a transition metal-N-doped carbon was reported as a functional unit for rational design of programmable functional reaction networks that exhibit dynamic behavior. We demonstrated that a network built around feedback and feedforward mechanism, along with other functions, such as deceleration, anaplerotic reaction, and cascade reaction by linking multiple network modules in microfluidic flow reactors. Interestingly, as a result, selective recognition of reductive biological small molecules was demonstrated using the proposed network. The methodology developed here provided a general framework to construct dissipative, tunable and robust (bio)chemical reaction networks, and extended biological applications of life-like materials.