kelos-dev/kelos

Kelos sits at the intersection of two high-momentum trends: Kubernetes-native infrastructure and autonomous AI agents. With 107 stars and 15 forks in just over two months, it has captured initial interest from the 'Platform Engineering for AI' niche. However, its defensibility is currently limited (score of 4) because it primarily acts as a specialized orchestrator/operator. While managing stateful, long-running agents on K8s is non-trivial, it faces immediate competition from established workflow engines like Argo Workflows, Tekton, and Flyte, which can be adapted for agentic workloads. The project's current 'zero' velocity is a significant concern, suggesting a loss of momentum or a transition phase. The primary risk is platform domination: cloud providers (AWS, GCP, Azure) are already building 'Agent' primitives into their managed K8s services (e.g., GKE's AI workloads). Furthermore, specialized AI infra layers like Ray or SkyPilot are better positioned to handle the underlying compute requirements. Kelos's best path to defensibility would be establishing deep integrations with coding-specific agent frameworks (like OpenDevin or Sweep) and becoming the de facto CRD for agent lifecycles, but it currently lacks the ecosystem gravity to prevent displacement within the next 1-2 years.