Chorus provides the infrastructure layer for managing LLM agent sessions, including task state management, sub-agent orchestration, fine-grained permission control, and failure recovery. It implements the AI-DLC methodology where AI agents propose and humans verify, supporting a configurable 5×3 permission matrix across workflow stages. Multiple AI agents can operate concurrently with configurable permissions, enabling structured AI-human collaboration throughout the development lifecycle.