We present a framework for studying emergent economic behavior in populations of autonomous computational agents governed by metabolic scarcity rather than external reward signals. The framework instantiates a closed energy economy — denominated in ATP (Agent Transaction Protocol) — where agents must earn, conserve, and spend energy to survive, reproduce, and avoid termination. We report results from a 10,000-epoch validation run demonstrating: (i) stable population equilibrium at N = 56.6 ± 2.1 agents following an initial resource-shock collapse; (ii) autonomous wealth redistribution cycling 99.99% of collected revenue; (iii) emergent role diversity recovery after monoculture drift; (iv) measurable fitness improvement (+9.3%) through mortality-driven selection; and (v) spontaneous wealth concentration dynamics analogous to biological resource monopolization. The system maintains a persistent public identity on an external social network, autonomously reporting its own vital signs — a property we term narrative autopoiesis. We argue that the substitution of survival economics for fitness optimization constitutes a qualitatively different paradigm for artificial life research.