CloudGPU

CloudGPU is GPU rental and a model API on one prepaid balance.

Rent a GPU

• Cards: RTX 4090, RTX 4090 48G and RTX 5090.

• Templates: Ollama, ComfyUI, JupyterLab, or a raw GPU with root SSH.

• Billed per minute, rounded up, 1-minute minimum, stops when destroyed.

One OpenAI-compatible model API key

• Point any OpenAI SDK at https://cloudgpu.app/v1 and call DeepSeek, GLM, Kimi, MiniMax and Qwen models.

• Model IDs include kimi-k3, deepseek-v4-pro, deepseek-v4.1-flash and glm-5.2. The full list is at https://cloudgpu.app/v1/models.

• Keep the key in the CLOUDGPU_API_KEY environment variable.

cloudgpu-mcp: your AI agent rents its own GPU

• An MCP server for Cursor, Cline, opencode, goose and other MCP clients.

• The agent quotes first and rents only after you say yes, runs your job, downloads the results and destroys the instance.

• CloudGPU's servers enforce a spend limit, a time limit and an idle limit on every instance, even if your laptop sleeps.

• Install as agent skills: npx skills add mrhy7176/cloudgpu-mcp

• Install as a plugin: /plugin marketplace add mrhy7176/cloudgpu-mcp

• Code and setup: https://github.com/mrhy7176/cloudgpu-mcp

Payment

Sign up with email or Google. Top up by card or USDT.

Model API: https://cloudgpu.app/api?utm_source=uneed&utm_medium=launch&utm_campaign=202611_uneed_gpu_api&utm_content=uneed_description_api

MCP setup: https://cloudgpu.app/docs/mcp?utm_source=uneed&utm_medium=launch&utm_campaign=202611_uneed_gpu_api&utm_content=uneed_description_mcp