# CloudGPU MCP

> CloudGPU MCP is a paid MCP server for cloud and infrastructure. Lets coding assistants rent GPUs by the minute, run jobs and shut them down within budgets. It can run locally on your machine or as a hosted remote server.

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

## About

The server gives an assistant a full rental loop: quote a price, rent once you approve, upload code and data, run the job, download results, then destroy the instance. Budgets and time limits are enforced on CloudGPU's servers, so a GPU still shuts down if the chat ends or a laptop sleeps. Billing is by the minute on selected options, while whole-hour options charge each started hour.

An agent can start without a key, using public discovery and `connect_account` to get a scoped key you approve in the browser with a spending limit. `request_top_up` only creates a link for the human owner and never completes payment. A `run_job` tool runs one command start to finish on CloudGPU's servers, and retrying it never starts a second GPU.

The page lists hourly rates, for example RTX 4090 24G at $0.39 and A100 80G at $1.59, and says it is in public beta with no SLA.

## Features

- **Quote before renting:** Live price and balance forecast before any GPU is created.
- **Server-side limits:** Budget and time limits shut GPUs down even if the client disconnects.
- **Retry-safe jobs:** run_job never starts a second GPU when retried.
- **Scoped agent keys:** Owner approves an agent's key with a spending limit.
- **Templates:** JupyterLab, Ollama, ComfyUI and LLaMA-Factory images.

## Good for

- Fine-tune a model on a rented GPU
- Run a ComfyUI or Ollama environment
- Batch a one-off compute job from a chat
- Compare GPU prices before renting

## Install

### Claude Code (Terminal)

```sh
claude mcp add --transport http cloudgpu-mcp https://cloudgpu.app/mcp
```

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

```sh
https://cloudgpu.app/mcp
```

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

```json
{
  "mcpServers": {
    "cloudgpu-mcp": {
      "url": "https://cloudgpu.app/mcp"
    }
  }
}
```

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

```json
{
  "servers": {
    "cloudgpu-mcp": {
      "type": "http",
      "url": "https://cloudgpu.app/mcp"
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "cloudgpu-mcp": {
      "serverUrl": "https://cloudgpu.app/mcp"
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "cloudgpu-mcp": {
      "type": "http",
      "url": "https://cloudgpu.app/mcp"
    }
  }
}
```

## Tools

- `get_account`: Balance and instance limit.
- `list_gpus`: GPU models with VRAM, price and availability.
- `list_templates`: Ready-made images such as JupyterLab and Ollama.
- `estimate_cost`: Live quote for a GPU and hours.
- `create_instance`: Rent a GPU from an approved quote.
- `run_command`: Run a command on the instance.
- `tail_logs`: Latest log lines and GPU utilisation.
- `upload`: Copy files to the instance.
- `download`: Copy files from the instance.
- `stop_instance`: Destroy the instance and stop billing.
- `run_job`: Run one command on a GPU start to finish.
- `connect_account`: Get an owner-approved scoped key.

## FAQ

### Does the account balance expire?

The pricing page says the balance does not expire.

### Is there an SLA?

No. CloudGPU is in public beta with no uptime guarantee.

### Does the assistant pay for itself?

No. request_top_up only makes a link for the human owner.

### What do I need installed?

Node.js 18 or newer.

### Is CloudGPU MCP free?

No. CloudGPU MCP is a paid product.

### How do I add CloudGPU MCP to Claude Code?

Run this command in your terminal: claude mcp add --transport http cloudgpu-mcp https://cloudgpu.app/mcp

### What are the best alternatives to CloudGPU MCP?

The closest alternatives to CloudGPU MCP 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/cloudgpu-mcp
