A podcast episode exploring kagent, an open-source Kubernetes operator for running AI agents and MCP (Model Context Protocol) tools on Kubernetes. The conversation covers what agents actually are (system prompt + LLM + tools), how MCP standardizes tool interfaces, the challenges of running agentic workloads in production, security and identity management gaps in the MCP spec, agent-to-agent architectures resembling microservices, and the cost implications of LLM API calls. A live demo shows creating a Kubernetes and GitHub agent using kagent with a Helm install.
ā¢46m watch time
9 Impressions