Agent Versioning

The practice of managing distinct versions of an AI agent, including its system prompt, tool set, model configuration, and associated evidence receipts, enabling rollback, A/B testing, and controlled feature rollouts without disrupting active deployments.