Build Optimized Supply Chains with NVIDIA cuOpt Agents and Skills

Learn how to run the NVIDIA cuOpt Supply Chain Agent — an open-source reference architecture inspired by NVIDIA's own AI-first supply chain planner, built for global-scale logistics optimization. In this tutorial, you'll deploy the agent using the BREV launchable, which sets up a preconfigured environment with cuOpt and all required dependencies. From there, the Jupyter notebook walks through initializing the agent, loading supply chain scenarios, and executing GPU-accelerated optimization workflows. NVIDIA cuOpt, an open-source GPU-accelerated optimization engine powers this multi-agent architecture to convert natural language problem descriptions into formal LP and MIP models — covering network design, fulfillment sequencing, and commitment planning. Using LangChain Deep Agent workflows for multi-agent orchestration, it runs what-if scenarios, compares against baselines, and returns optimized decisions with clear explainability. No manual constraint tuning. Modular cuOpt agent skills — including Customer Allocation, Generic Max Supply, and cuOpt Solver — let developers inject custom constraints, integrate proprietary data, and expand into broader MILP domains. The system is fully extensible. Explore the cuOpt agent skills on GitHub and start building for your own enterprise supply chain workflows. Build using this BREV launchable and start optimizing: https://brev.nvidia.com/launchable/deploy/now?launchableID=env-3CKDISJPNJQCgbvH6WPgxCxU6FF Free developer credits applied to the first 50 users who deploy the launchable. 📥 Get Started: https://github.com/NVIDIA/cuopt/blob/main/AGENTS.md ➡️ Learn more: https://github.com/NVIDIA/cuopt-examples/tree/main/cuopt-agent 💡Brev Lauchable: https://nvda.ws/42LU7Mz 📝 Tech blog: https://nvda.ws/3QRZJ5m