Kynara is the permission control plane for AI agents. Enforce RBAC + ABAC policies, require human approval for sensitive actions, and maintain a tamper-evident audit log — for LangChain, AutoGen, CrewAI, OpenAI, and Anthropic agents.