Agyn: open-source platform for scalable, secure AI agents on Kubernetes

Researchers Nikita Benkovich and Vitalii Valkov have released Agyn, an open-source platform addressing the operational challenges of deploying AI agents in production. Unlike simple chatbots, AI agents execute non-deterministic workflows, maintain stateful sessions, and often require privileged access to internal services. Agyn tackles these complexities with three core principles: a signal-driven, stateful serverless runtime built on Kubernetes; a Terraform provider that enables agent and harness definition as code; and a security model grounded in zero-trust and least-privilege principles. The platform is agent-agnostic, model-agnostic, and cloud-agnostic, allowing organizations to run any agent framework on any infrastructure.

Designed for enterprise-scale operations, Agyn abstracts infrastructure management while ensuring proper isolation between agent instances. Its serverless runtime automatically scales based on incoming signals, and the stateful design allows agents to maintain context across long-running sessions. The Terraform integration means DevOps teams can version-control agent configurations, making deployments reproducible and auditable. Meanwhile, the zero-trust security model ensures each agent receives only the minimum permissions needed, reducing blast radius in case of compromise. The paper (arXiv:2605.27575) details the architecture and benchmarks showing efficient resource utilization. For organizations moving from prototype AI agents to production systems, Agyn provides a ready-made foundation for governance and operations.