About Nvidia MCP
Nvidia MCP is an MCP server in the RAG category: mCP server that gives Claude, Cursor and any AI agent free access to 100+ NVIDIA-hosted models (Nemotron, Llama, GPT-OSS, DeepSeek, vision, embeddings) with automatic task routing and parallel multi-model fan-out.
Install
git clone https://github.com/holm-digital-io/nvidia-mcpThis server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.
Tools (10)
nv_autoRoutes and runs in one step. Pass an image and it selects a vision model.
Parameters not published.
nv_callCalls one specific model. Supports images for vision models.
Parameters not published.
nv_categoriesLists every task category the router knows, with its top models.
Parameters not published.
nv_embedEmbeds texts. Returns shape only unless you pass `save_to`, to save context.
Parameters not published.
nv_fanoutRuns one prompt across up to 6 models **in parallel**, returns all answers.
Parameters not published.
nv_healthVerifies the API key and reports catalog size.
Parameters not published.
nv_list_modelsLive model list from your account, tagged by capability and liveness.
Parameters not published.
nv_probeActually calls each model and reports which ones answer, and how fast.
Parameters not published.
nv_route"What should I use for X?" Recommends models without calling any.
Parameters not published.
nv_similarityRanks candidate strings against a query by cosine similarity.
Parameters not published.
Ask AI
Ask AI about Nvidia MCP
Powered by Claude · Grounded in docs
Security checks
6 checks have not been run yet.