# gpu.nz GPU cloud

> gpu.nz GPU cloud is a paid MCP server for cloud and infrastructure. Search GPU offers, estimate VRAM, and rent or stop cloud GPU servers from an agent. It can run locally on your machine or as a hosted remote server.

- **Kind:** MCP server
- **Pricing:** Paid
- **Platforms:** API, CLI
- **Website:** https://gpu.nz/en/docs/mcp
- **Repository:** https://github.com/gpunz/gpunz-mcp
- **MCP Registry name:** nz.gpu/gpu-cloud
- **Categories:** cloud and infrastructure, AI models and LLM tooling
- **Listing updated:** 2026-11-14

## About

Browsing works anonymously; spending needs a key. Tools such as list_gpu_models, search_offers and estimate_vram need no credentials, so an assistant can compare live per-hour prices, filter by country or reliability and work out how much memory a model like Qwen, DeepSeek or FLUX needs at a given quantisation and context. Account tools need a gpz_ API key whose scopes you choose.

With a key the agent can create an instance from a template (ComfyUI, Forge, Ollama, vLLM, Jupyter, LoRA training or a custom Docker image), read its boot logs, stop it so only the disk is billed, restart it, or destroy it along with its disk. Billing starts the moment a server is created, so the server flags renting, starting and destroying as consequential and clients ask for confirmation.

The agent cannot spend more than the account balance, because top-ups happen only in the browser through a link from topup_link. The limit is 120 requests a minute per key and rent, stop and destroy actions go to the account audit log. The server is hosted at gpu.nz/mcp, with a Docker or npx stdio bridge for local-only clients.

## Features

- **Keyless price search:** Lists GPU models and live offers with filters for count, memory, price, country and reliability.
- **VRAM estimator:** Estimates memory needs for named models and suggests the cheapest GPUs that fit.
- **Ready templates:** Launches ComfyUI, Forge, Ollama with Open WebUI, vLLM, Jupyter, Kohya, FluxGym, Unsloth or your own image.
- **Instance lifecycle:** Creates, stops, starts and destroys servers and reads their recent logs.
- **Scoped API keys:** Key scopes limit what an agent may read or write, with expiry and revocation.
- **Spend guardrails:** Spending is capped by balance and top-ups happen in the browser only.

## Good for

- Find the cheapest RTX 4090 above a reliability threshold
- Work out whether a 27B model fits a given GPU at 32k context
- Start a ComfyUI or vLLM server and get its access link
- Check running servers and stop idle ones
- Get a balance top-up link without leaving the chat

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http gpu-cloud https://gpu.nz/mcp
```

### Claude Desktop (Settings → Connectors → Add custom connector)

```sh
https://gpu.nz/mcp
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "gpu-cloud": {
      "url": "https://gpu.nz/mcp"
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "gpu-cloud": {
      "type": "http",
      "url": "https://gpu.nz/mcp"
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "gpu-cloud": {
      "serverUrl": "https://gpu.nz/mcp"
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "gpu-cloud": {
      "type": "http",
      "url": "https://gpu.nz/mcp"
    }
  }
}
```

## Tools

- `list_gpu_models`: GPU models in stock with memory, class and price per hour.
- `search_offers`: Servers with prices, filtered by model, GPU count, memory, price, country and reliability.
- `get_offer`: One server and its current price.
- `list_templates`: Ready environments such as ComfyUI, Forge, Ollama, vLLM and Jupyter.
- `estimate_vram`: GPU memory a model needs and which GPUs fit, by price.
- `topup_link`: A browser link to top up the balance.
- `get_account`: Account, balance, hourly spend and runway.
- `list_instances`: Your servers with status, GPU, price and SSH command.
- `get_instance`: One server with SSH, app links and cost since start.
- `get_instance_logs`: The last log lines covering boot, app start-up and errors.
- `create_instance`: Rents a server from a template; billing starts immediately.
- `stop_instance`: Stops a server so only its disk is billed.
- `start_instance`: Starts a stopped server.
- `destroy_instance`: Destroys a server and its disk irreversibly.
- `list_ssh_keys`: The account's SSH keys.
- `add_ssh_key`: Adds an SSH key to the account.

## FAQ

### Do I need an API key?

Not for GPU search, prices, templates or VRAM estimates. Renting and managing servers requires a key starting gpz_.

### Can the agent overspend?

No. It cannot spend more than the balance holds, and top-ups are completed in the browser only.

### What does stopping a server cost?

Stopping ends GPU billing, but the disk stays billed while it exists, according to the pricing page.

### How is rental billed?

Per minute from the account balance, with the price fixed at launch; traffic is billed per gigabyte.

### Is destroying a server reversible?

No. destroy_instance removes the server and its disk irreversibly.

### Is gpu.nz GPU cloud free?

No. gpu.nz GPU cloud is a paid product.

### How do I add gpu.nz GPU cloud to Claude Code?

Run this command in your terminal: claude mcp add --transport http gpu-cloud https://gpu.nz/mcp

### What are the best alternatives to gpu.nz GPU cloud?

The closest alternatives to gpu.nz GPU cloud are Claude Code Ultimate Guide, Agent Yurt, Proofline, signafAIde and Toolkien.

## Alternatives

- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.
- [Agent Yurt](https://ai-powered.com/mcp/agent-yurt): Agent Yurt is a free MCP server for knowledge and memory.
- [Proofline](https://ai-powered.com/mcp/proofline): Proofline is a freemium MCP server for developer tools.
- [signafAIde](https://ai-powered.com/mcp/signafaide): signafAIde is a freemium MCP server for security.
- [Toolkien](https://ai-powered.com/mcp/toolkien): Toolkien is an open-source MCP server for ai models and llm tooling.
- [filex](https://ai-powered.com/mcp/filex): filex is an open-source MCP server for files and documents.
- [Lugh](https://ai-powered.com/mcp/lugh): Lugh is an open-source MCP server for ai models and llm tooling.
- [HypePaper](https://ai-powered.com/mcp/hypepaper): HypePaper is a freemium MCP server for ai models and llm tooling.

---
Source: https://ai-powered.com/mcp/gpu-nz-gpu-cloud
