Kimi AI and kvcache-ai have made their AgentENV system, which supports agentic reinforcement learning training, available under the MIT license. This open-source release provides a distributed system that can run agent sandboxes within Firecracker microVMs, offering features such as millisecond snapshot and resume capabilities. The system is designed to work behind an E2B-compatible API. This move is significant as it allows developers to access and build upon the technology, potentially accelerating advancements in reinforcement learning.