Nvidia Nim MCP — Communication skill for Claude Code

nvidia-nim-mcp

An MCP ([Model Context Protocol](https://modelcontextprotocol.io)) server that plugs [NVIDIA NIM](https://build.nvidia.com/)'s free-tier models straight into Claude Code — image generation, translation, LLM chat, vision, content safety, embeddings, and a provider health check — behind seven small, consistent tools.

The idea is simple: **a model being slow, rate-limited, or quietly retired should never take a tool down.** Every capability tries more than one model, and two of them — translation and the "ask another LLM" tool — keep going past NVIDIA into whichever free-tier providers you've configured (Groq, Mistral, Gemini, Cerebras). The caller never has to know or care which model actually answered.

Zero cost. No credit card. Just an API key from [build.nvidia.com](https://build.nvidia.com/).

Table of Contents

• Tools
• The fallback chain
• Setup
• Example usage
• Output files
• Development
• Project layout
• License

🧰 Tools


Seven tools. Six are backed by their own model chain; the seventh checks the health of every model the other six actually use. Every generation tool now has at least one fallback beyond NVIDIA — see [The fallback chain](#-the-fallback-chain) for exactly how each one is shaped.

Tool	What it does	NVIDIA model chain (in order)	Fallback beyond NVIDIA
🖼️ `generat