
Lets your coding agent look up NVIDIA's current CUDA documentation and code examples, served by NVIDIA itself.
by Nvidia· Developer toolsUnclaimed · Yours? Claim it
4 images
About
NVIDIA's hosted CUDA MCP Server, part of Nsight AI. It connects your AI coding agent to current, first-party NVIDIA CUDA documentation and code examples, so the answers your agent gives about CUDA draw on NVIDIA's own docs. What you can do with it - Search NVIDIA's CUDA documentation from your agent - Find NVIDIA's CUDA code examples to build from - Ask CUDA questions while you write, profile and improve your code, and get answers grounded in NVIDIA-curated docs How it connects NVIDIA hosts it at https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs. The first time you connect, you sign in with your NVIDIA Developer account, and your agent reuses that sign-in afterwards. NVIDIA gives setup steps for Claude Code, Codex, Cursor and Antigravity, and any other agent that takes a hosted MCP server address can connect to it. Good to know - It's NVIDIA's suggested route for people moving off the Nsight Copilot extension for Visual Studio Code, which NVIDIA is deprecating. - The hosted server serves NVIDIA-curated documentation. If you work with highly sensitive or proprietary code, NVIDIA suggests its open-source, self-hosted Nsight Copilot Blueprint instead, so your data stays on your own systems. - It complements your usual debugging and profiling tools rather than replacing them.
Plugins people install alongside NVIDIA CUDA Docs MCP.
Once it’s added
“Search the CUDA docs for how to avoid uncoalesced memory accesses.”
“Find a CUDA code example of matrix multiplication with shared memory.”
“How do I overlap data transfers and kernels with CUDA streams? Check NVIDIA's docs.”
“Look up what the CUDA documentation says about cooperative groups.”
“Check NVIDIA's current CUDA docs before we use this API in our kernel.”
Before you add it
Reads only
NVIDIA's CUDA documentation · NVIDIA's CUDA code examples
From people who use it
Step by step
Nothing new reaches your agent until a person has looked at it.
Scanned on every change, and again when our rules improve.
A hosted server: nothing runs on your machine.
A hosted server: nothing runs on your machine.
A hosted server: nothing runs on your machine.
A hosted server: nothing runs on your machine.
A hosted server: nothing runs on your machine.
A hosted server: nothing runs on your machine.
A hosted server: nothing runs on your machine.
Add to VS Codecopilot mcp add --transport http nvidia-cuda-docs-mcp https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs
A hosted server: nothing runs on your machine.
Add to Cursor.cursor/mcp.json{
"mcpServers": {
"nvidia-cuda-docs-mcp": {
"url": "https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs"
}
}
}A hosted server: nothing runs on your machine.
claude mcp add --transport http nvidia-cuda-docs-mcp https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs
A hosted server: nothing runs on your machine.
codex mcp add nvidia-cuda-docs-mcp --url https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs
A hosted server: nothing runs on your machine.
~/.gemini/config/mcp_config.json{
"mcpServers": {
"nvidia-cuda-docs-mcp": {
"url": "https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs"
}
}
}A hosted server: nothing runs on your machine.
gemini mcp add --transport http nvidia-cuda-docs-mcp https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs
A hosted server: nothing runs on your machine.
A hosted server: nothing runs on your machine.
openclaw mcp add nvidia-cuda-docs-mcp --url https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs --transport streamable-http
A hosted server: nothing runs on your machine.
opencode mcp add nvidia-cuda-docs-mcp --url https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs
A hosted server: nothing runs on your machine.
hermes mcp add nvidia-cuda-docs-mcp --url https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs
~/.hermes/config.yaml{
"mcp_servers": {
"nvidia-cuda-docs-mcp": {
"url": "https://api.copilot.nsight.ngc.nvidia.com/mcp/cuda-docs"
}
}
}Scanned and passed · checked 8 h ago
No reviews yet. Used NVIDIA CUDA Docs MCP? Say how it went.
Write a reviewPlugins people install alongside NVIDIA CUDA Docs MCP.
Matt Pocock's everyday engineering skills: grilling sessions, specs and tickets, TDD, bug diagnosis, code review and writing for agents.
Addy Osmani's engineering workflows for coding agents: spec, plan, build, test, review and ship, with quality gates at each step.
Nobody has shared how they use NVIDIA CUDA Docs MCP yet.
Security scan
Passed
Safe to install in every agent it supports. Runs remotely on api.copilot.nsight.ngc.nvidia.com · needs your sign-in.