Abstract: Stateless Swarm Agent Architecture for Scalable, Privacy-Aligned AI This paper proposes a novel agentic architecture rooted in swarm principles, designed to enable scalable, stateless, and ephemeral computation for distributed AI systems. Diverging from traditional multi-agent paradigms that rely on persistent state, inter-agent communication, or shared memory contexts, we introduce a model in which agents function as deterministic, testable microservices. Each agent is instantiated transiently to perform a bounded computation