NVIDIANVIDIAOfficial·6 Agent Skills Included

cuopt

Solve routing, LP, MILP, and QP problems on GPU

Solves vehicle routing, linear, mixed-integer, and quadratic optimization problems using NVIDIA GPU acceleration. Removes the guesswork from formulating models, installing the right CUDA-matched packages, and calling Python, C, CLI, or REST APIs correctly. Guides agents to clarify ambiguous requirements, verify results, and explore multi-objective tradeoffs with Pareto frontiers.
npx skills add NVIDIA/cuopt --all -g -y
Available:

Routes the AI agent to the correct cuOpt skill by task and interface, enforces asking when problems are ambiguous, and requires proposing skill updates after corrections or discovered gotchas.

All Skills in This Repository (6)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install cuOpt?

Run `npx skills add NVIDIA/cuopt --all -g -y` in your terminal to install all cuOpt agent skills globally.

What problems can NVIDIA cuOpt solve?

cuOpt solves vehicle routing (TSP, VRP, PDP), linear programming, mixed-integer programming, and quadratic programming problems at GPU speed. It handles large-scale models with millions of variables and constraints.

How do I solve a vehicle routing problem with cuOpt?

The routing skill walks your agent through building a DataModel with cost matrices, time windows, and fleet capacities, then calling routing.Solve in Python. It also covers checking solution status and debugging infeasible orders.

Does cuOpt work with Claude Code and other AI agents?

Yes. All skills follow the standard SKILL.md format and work in Claude Code, Codex, Cursor, and other compatible agents, with an AGENTS.md file that routes requests to the right skill.

Do I need a GPU to use cuOpt?

Yes. cuOpt requires an NVIDIA GPU with Compute Capability 7.0 or newer (Volta or later) and CUDA 12.x or 13.x. The install skill helps match the correct package to your CUDA version.

Related Repositories in Data & Analytics

View All in Data & Analytics