Docs and knowledge - official NVIDIA server
NVIDIA CUDA MCP server
NVIDIA-hosted remote MCP server that connects AI coding agents to current, first-party NVIDIA CUDA documentation and code examples. Part of NVIDIA Nsight AI. NVIDIA hosts it at https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs with OAuth auth.
Last verified 1 tool2 sources
Install
Claude Code
About Claude Codeclaude mcp add --scope user --transport http nvidia-cuda-docs https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docsCodex
About Codexcodex mcp add nvidia-cuda-docs --url https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docsCursor
About Cursor{
"mcpServers": {
"nvidia-cuda-docs": {
"url": "https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs"
}
}
}Generic
https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docsTools
search_cuda_docs- Semantic search over a curated corpus of NVIDIA CUDA documentation and code samples (name documented for the self-hosted Nsight Copilot Blueprint endpoint at the same /mcp/cuda-docs path).
As documented on October 17, 2026. A running server reports its current list through tools/list.
Notes and caveats
- First connection requires signing in with an NVIDIA Developer account; the client reuses that authentication afterward.
- NVIDIA recommends the open-source, self-hosted Nsight Copilot Blueprint (github.com/NVIDIA-AI-Blueprints/nsight-copilot, exposes the same /mcp/cuda-docs endpoint locally) for sensitive or proprietary code.
- The Nsight Copilot VS Code extension is being deprecated in favor of this MCP server.
- NVIDIA also documents an Antigravity config using a serverUrl key.
More MCP servers for docs and knowledge
- NVIDIA
Isaac Sim MCP server
Self-hosted MCP server that helps AI coding assistants with NVIDIA Isaac Sim development by searching Isaac Sim extensions, APIs, code examples, documentation and configuration settings.
Transport: Streamable HTTPAuth: API key - NVIDIA
Kit MCP server
Self-hosted MCP server that helps AI coding assistants with NVIDIA Omniverse Kit development: searching Kit extensions, extension APIs and dependencies, code and test examples, settings and documentation.
Transport: Streamable HTTPAuth: API key - NVIDIA
NVIDIA Elements MCP server
Local MCP server started with `nve mcp` from the NVIDIA Elements CLI, exposing Elements design system component APIs, examples, design tokens, package info and project setup tools to AI assistants.
Transport: stdioAuth: No auth - NVIDIA
OmniUI MCP server
Self-hosted MCP server that gives AI coding assistants omni.ui knowledge for NVIDIA Omniverse: UI classes, modules and methods, widget docs, styling docs and code and window-layout examples.
Transport: Streamable HTTPAuth: API key - NVIDIA
USD Code MCP server
Self-hosted MCP server from NVIDIA Omniverse that gives AI coding assistants USD/OpenUSD API knowledge (modules, classes, methods, code examples) via semantic search and NVIDIA reranking.
Transport: Streamable HTTPAuth: API key - Adobe
React Aria MCP server
Local MCP server run via npx that lets IDE agents list React Aria documentation pages and fetch full markdown for components, hooks and guides.
Transport: stdioAuth: No auth
Related reading
- RTX Spark for Local LLMs: Will It Run a Coding Agent?RTX Spark laptops ship October 16 with up to 128 GB of unified memory. Why the 600 GB/s figure is not memory bandwidth and which local coding models fit.
- NVIDIA OpenShell Makes Agent Sandboxes a Policy LayerNVIDIA OpenShell runs autonomous agents in policy-enforced sandboxes: an open-source runtime that moves agent security from prompt rules to infrastructure.
- AMD MI355X vs NVIDIA B200 vs B300 for Open-Weight ServingAMD MI355X vs NVIDIA B200 vs B300 for serving Kimi K3's about 1.5TB of open weights: verified specs, the Wafer benchmark, and cost per token for each.
- Nvidia, Microsoft, Meta, and 30+ Companies Warn Against Overregulating Open-Weight AI ModelsAn open letter from Nvidia, Microsoft, Meta, OpenAI and 30+ signers defends open-weight AI models. What it argues, who stayed out, and how HN reacted.
- NVIDIA Nemotron 3 Super: A Developer's Guide to the 120B Hybrid MoENemotron 3 Super developer guide: the latent MoE hybrid Mamba architecture, hardware needs, reasoning modes, and code to run NVIDIA's 120B model.
- NVIDIA's Nemotron 3 Super in 6 MinutesNVIDIA Nemotron 3 Super combines latent mixture of experts with hybrid Mamba layers - 120B total parameters, 12B active per token, 1M context and open weights.
FAQ
What does the NVIDIA CUDA MCP server do?
NVIDIA-hosted remote MCP server that connects AI coding agents to current, first-party NVIDIA CUDA documentation and code examples. Part of NVIDIA Nsight AI. The vendor documents 1 tool, including search_cuda_docs.
How do I add the NVIDIA CUDA MCP server to Claude Code?
Run claude mcp add --scope user --transport http nvidia-cuda-docs https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs, copied from NVIDIA's docs. This page also has snippets for Codex, Cursor and Generic.
Does the NVIDIA CUDA MCP server need authentication?
Yes, OAuth. Transport: Streamable HTTP. Status: Unlabelled, last checked October 17, 2026.
Sources
Last verified against 2 NVIDIA pages, then re-checked independently.
Newsletter
New MCP servers, the week they ship
One email per week. New official MCP servers, the coding tools that support them and how developers wire them up.