Open-source authentication for AI agents is an interesting direction, especially as more applications need to handle agent access alongside traditional users. I’m particularly interested in how the new version handles permissions and keeps authentication manageable as these systems become more complex.