Moonshot AI and kvcache-ai Open-Source AgentENV for Scalable Agent Workflows

By Alexis Dufresne · Published · Updated ## TL;DR Moonshot's Kimi team and kvcache-ai open-sourced AgentENV, the Firecracker-based platform they say powered agentic RL training for the 2.8-trillion-parameter Kimi K3. Each microVM boots or resumes in under 50 ms and pauses in under 100 ms, with incremental snapshots completing in under 100 ms. A running sandbox can fork into up to 16 independent children on the same node, and the API is drop-in compatible with the E2B SDK. [...] Sung Kim @sungkim.bsky.social Moonshot AI's / Kimi's AgentENV (AENV) is a platform for running agent environments at scale, powering agentic RL training for Kimi K3. github.com/kvcache-ai/A... View on Bluesky → Sung Kim @sungkim.bsky.social: Moonshot AI's / Kimi's AgentENV (AENV) is a platform for running agent environments at scale, powering agentic RL training for Kimi K3. git… → Originally reported by github.com Read the original article → Original headline: GitHub - kvcache-ai/AgentENV: Age

Kimi AI and kvcache-ai Open Sources 'AgentENV': A Distributed System that Powers Agentic Reinforcement Learning (RL) Training for Kimi K3 [...] '/%3E%3Cpath d='M16.0001 7.9996c0 4.418-3.5815 7.9996-7.9995 7.9996S.001 12.4176.001 7.9996 3.5825 0 8.0006 0C12.4186 0 16 3.5815 16 7.9996Z' fill='url(%23paint1_radial_15251_63610)'/%3E%3Cpath d='M16.0001 7.9996c0 4.418-3.5815 7.9996-7.9995 7.9996S.001 12.4176.001 7.9996 3.5825 0 8.0006 0C12.4186 0 16 3.5815 16 7.9996Z' fill='url(%23paint2_radial_15251_63610)' fill-opacity='.5'/%3E%3Cpath d='M7.3014 3.8662a.6974.6974 0 0 1 .6974-.6977c.6742 0 1.2207.5465 1.2207 1.2206v1.7464a.101.101 0 0 0 .101.101h1.7953c.992 0 1.7232.9273 1.4917 1.892l-.4572 1.9047a2.301 2.301 0 0 1-2.2374 1.764H6.9185a.5752.5752 0 0 1-.5752-.5752V7.7384c0-.4168.097-.8278.2834-1.2005l.2856-.5712a3.6878 3.6878 0 0 0 .3893-1.6509l-.0002-.4496ZM4.367 7a.767.767 0 0 0-.7669.767v3.2598a.767.767 0 0 0 [...] ## Marktechpost's Post ### Marktechpost

Follow aimomentumnews's profile picture aimomentumnews 𝗞𝗶𝗺𝗶 𝗔𝗜 𝗮𝗻𝗱 𝗸𝘃𝗰𝗮𝗰𝗵𝗲-𝗮𝗶 𝗢𝗽𝗲𝗻 𝗦𝗼𝘂𝗿𝗰𝗲𝘀 ‘𝗔𝗴𝗲𝗻𝘁𝗘𝗡𝗩’: 𝗔 𝗗𝗶𝘀𝘁𝗿𝗶𝗯𝘂𝘁𝗲𝗱 𝗦𝘆𝘀𝘁𝗲𝗺 𝘁𝗵𝗮𝘁 𝗣𝗼𝘄𝗲𝗿𝘀 𝗔𝗴𝗲𝗻𝘁𝗶𝗰 𝗥𝗲𝗶𝗻𝗳𝗼𝗿𝗰𝗲𝗺𝗲𝗻𝘁 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 (𝗥𝗟) 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗳𝗼𝗿 𝗞𝗶𝗺𝗶 𝗞𝟯 Moonshot AI's Kimi team and kvcache-ai open-sourced AgentENV (AENV) under MIT, as part of Kimi K3 Open Day. It runs agent sandboxes as Firecracker microVMs with millisecond snapshot, resume, and 16-way fork, behind an E2B-compatible API. The post Kimi AI and kvcache-ai Open Sources ‘AgentENV’: A Distributed System that Powers Agentic Reinforcement Learning (RL) Training for Kimi K3 appeared first on MarkTechPost. let us know what you think below ⬇️ 🌎 #AI #World #News #ArtificialIntelligence #AIWorldNewsToday ## No comments yet. Start the conversation. Log in to like or comment. [...] Log In Sign Up aimomentumnews's profile picture Never miss a post from aimomentumnews

46,353 followers We've open-sourced AgentENV in collaboration with kvcache-ai. AgentENV is a distributed system for running agent environments at scale. Its components power agentic RL training for Kimi K3, with fast snapshot, resume, and fork support for large-scale parallel agent workflows. Explore on GitHub: GitHub - kvcache-ai/AgentENV: AgentENV (AENV) is a distributed platform for running agent environments at scale. George Leonard, graphic ... Wish Kimi published models more able to run on local MAC/M platform/MLX... in a realistic RAM size. played with it using their online interface, one of the older models and liked it. would love to try the new models locally. G Vijay Nadkarni, graphic [...] ## LinkedIn respects your privacy LinkedIn and 3rd parties use essential and non-essential cookies to provide, secure, analyze and improve our Services, and to show you relevant ads (including professional and job ads) on and off LinkedIn. Learn more in our Cookie Policy. Select A

Awesome!! Congras Kimi (Moonshot AI) To view or add a comment, sign in ## More Relevant Posts View profile for Sudhanshu Sharma ✨ AgentENV is best understood as part of a bigger shift in how frontier models are trained: not just better GPUs and bigger models, but better environment infrastructure for agentic reinforcement learning. Kudos Kimi (Moonshot AI) 👏 View organization page for Kimi (Moonshot AI) 46,353 followers We've open-sourced AgentENV in collaboration with kvcache-ai. AgentENV is a distributed system for running agent environments at scale. Its components power agentic RL training for Kimi K3, with fast snapshot, resume, and fork support for large-scale parallel agent workflows. Explore on GitHub: [...] Agree & Join LinkedIn By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy. # AgentENV Open-Sourced with kvcache-ai Collaboration View organization page for Kimi (Moonshot AI) 46,353 followers We've

# Kimi AI and kvcache-ai Open Sources ‘AgentENV’: A Distributed System that Powers Agentic Reinforcement Learning (RL) Training for Kimi K3 By Asif Razzaq - Moonshot AI’s Kimi team and kvcache-ai have open-sourced AgentENV (AENV), a distributed platform for running agent environments at scale. AgentENV powers agentic reinforcement learning (RL) training for Kimi K3, Moonshot’s 2.8-trillion-parameter Mixture-of-Experts model. The code ships under an MIT license. ## Why Environment Infra Holds Back Agentic RL Agentic RL does not just sample text. It requires the model to act inside a real computer. Every rollout needs an isolated Linux environment with a filesystem, a network stack, and live processes. [...] ## Snapshot, pause, resume, fork These four features are the reason the project exists. AgentENV snapshots memory and filesystem changes incrementally rather than writing a full image each time. The reported figures are as follows. Snapshot-backed environments boot or resume