Running AI MCP Tools on Kubernetes with kagent

In this episode of Agentic DevOps, Nirmal Mehta and his co-host, Bret Fisher, interview Eitan Yarmush from SoloIO about kagent, a Kubernetes tool operator for AI and agentic DevOps. They discuss the challenges and potential of using AI agents in infrastructure, focusing on the balance between excitement and apprehension due to the introduction of non-determinism. The conversation covers the definition of agents, the Model Context Protocol (MCP), and the importance of security and identity management in agentic workflows. They also explore the practical aspects of running and scaling AI agents, touching on the need for microservices architecture and the cost considerations of using LLMs, and kagent's role in deploying agents, while emphasizing that MCP servers are just applications with APIs.