gpu.nz GPU cloud

nz.gpu由 gpu.nz 發布官方發布者v1.0.0更新於 Oct 26, 2026

Rent cloud GPUs from AI agents: live prices, VRAM estimates, launch, logs and stop.

已驗證Streamable HTTP可網頁執行雲端與基礎設施金融AI 與機器學習

概覽

AI 產生的概覽

讓 AI 助理在 gpu.nz 上尋找、比價並租用雲端 GPU 伺服器，估算模型 VRAM、啟動範本、查看日誌並停止或銷毀伺服器。

可以做什麼
  • 瀏覽 GPU 型號，依即時價格搜尋可租用的伺服器並查看單一報價細節
  • 估算模型所需 VRAM 及適合的 GPU，並列出 ComfyUI、Ollama、vLLM、Jupyter 等現成範本
  • 依報價與範本租用伺服器，並啟動、停止或銷毀伺服器
  • 查看執行個體狀態、SSH 指令、應用程式網址、目前費用與最近的容器日誌
  • 查看帳戶餘額、每小時花費與可用時數，並管理 SSH 金鑰
  • 取得儲值連結，供使用者在瀏覽器中開啟
適用情境
適合讓助理先估算模型 VRAM、比較即時每小時價格，再實際租用、執行並關閉 GPU 伺服器，以完成影像生成、微調或推論等工作。
執行需求
可使用 gpu.nz 的遠端端點，或透過已發布的容器映像、Node.js 18+ 下的 npx 執行本機 stdio 橋接。瀏覽類工具不需憑證；租用與管理伺服器需要 gpu.nz 主控台的 API 金鑰，透過 GPUNZ_API_KEY 環境變數或 Authorization bearer 標頭傳入。instances:write、user:read、billing:read 等權限範圍決定可用工具。
安裝前請注意
租用後立即開始計費，按分鐘持續到停止或銷毀；destroy_instance 會永久刪除伺服器及其磁碟。API 金鑰可產生費用並存取帳戶，應妥善保管並限制權限。明確要求時會回傳 root 密碼。儲值只能在瀏覽器中完成，不會透過工具進行。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 gpu.nz GPU cloud，將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

交給代理程式安裝

把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。

閱讀 https://sourceweft.com/mcp/nz.gpu%2Fgpu-cloud/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。

檢視代理程式所依據的安裝指南

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

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

README

[gpu.nz MCP — let AI agents find a GPU, estimate VRAM, rent a server and stop it]

gpu.nz MCP server

Let Claude, Cursor, Codex and other AI agents rent cloud GPUs: search RTX 4090, A100, H100 and B200 servers with live prices, work out how much VRAM a model needs, launch ComfyUI, Ollama, vLLM or Jupyter, read the logs and stop the server when the job is done. Billed by the minute.

[MCP] [Endpoint] [License: MIT]

Docs · VRAM calculator · REST API · CLI · Русский

Connect

The server is hosted at https://gpu.nz/mcp — nothing to install. Browsing GPUs, prices and VRAM estimates works without a key; renting and managing servers needs an API key from gpu.nz/console/keys.

Claude Code

bash
claude mcp add --transport http gpunz https://gpu.nz/mcp \  --header "Authorization: Bearer gpz_…"

Cursor — ~/.cursor/mcp.json

json
{  "mcpServers": {    "gpunz": {      "url": "https://gpu.nz/mcp",      "headers": { "Authorization": "Bearer gpz_…" }    }  }}

VS Code — .vscode/mcp.json (asks for the key once and keeps it in secret storage)

json
{  "servers": {    "gpunz": {      "type": "http",      "url": "https://gpu.nz/mcp",      "headers": { "Authorization": "Bearer ${input:gpunz-key}" }    }  },  "inputs": [{ "type": "promptString", "id": "gpunz-key", "description": "gpu.nz API key", "password": true }]}

Codex CLI

bash
export GPUNZ_API_KEY="gpz_…"codex mcp add gpunz --url https://gpu.nz/mcp --bearer-token-env-var GPUNZ_API_KEY

Claude Desktop — claude_desktop_config.json, through mcp-remote

json
{  "mcpServers": {    "gpunz": {      "command": "npx",      "args": ["-y", "mcp-remote", "https://gpu.nz/mcp", "--header", "Authorization:${GPUNZ_AUTH}"],      "env": { "GPUNZ_AUTH": "Bearer gpz_…" }    }  }}

On claude.ai you can add https://gpu.nz/mcp under Connectors without a key: search and estimates work.

Local stdio bridge

For clients that only start local servers, this repository is a small stdio bridge to the hosted server:

bash
# Dockerdocker run -i --rm -e GPUNZ_API_KEY="gpz_…" ghcr.io/gpunz/gpunz-mcp
# Node.js 18+, straight from GitHubGPUNZ_API_KEY="gpz_…" npx -y github:gpunz/gpunz-mcp
Variable
GPUNZ_API_KEYAPI key (gpz_…). Optional: without it only the browsing tools work
GPUNZ_MCP_URLEndpoint override, default https://gpu.nz/mcp

Tools

ToolWhat it doesKey scope
list_gpu_modelsGPU models in stock: memory, class, cheapest price per hour—
search_offersServers with live prices; filter by GPU, count, VRAM, price, country, reliability—
get_offerOne server and its current price—
list_templatesReady environments: ComfyUI, Forge, Ollama + Open WebUI, vLLM, Jupyter, Kohya, FluxGym, Unsloth, your own Docker image—
estimate_vramVRAM a model needs (Qwen, DeepSeek, Llama, Gemma, GPT-OSS, FLUX, Wan…) at a quantisation and context, and the cheapest GPUs that fit—
topup_linkA link to top up the balance in the browser—
get_accountAccount, balance, hourly spend and runwayuser:read, billing:read
list_instances · get_instanceYour servers: status, SSH command, app URLs, cost so farinstances:read
get_instance_logsThe last log lines: boot, app start-up, errorsinstances:read
create_instanceRents a server with a template; billing starts at onceinstances:write
stop_instance · start_instanceStop (only the disk is billed) and start againinstances:write
destroy_instanceDestroys a server and its diskinstances:write
list_ssh_keys · add_ssh_keySSH keys installed on new serversuser:read / user:write

Try asking

  • How much VRAM does Qwen3.8 27B need with a 32k context, and where is it cheapest?
  • Find the cheapest RTX 4090 with reliability above 0.98 and start ComfyUI on it with an 80 GB disk.
  • Run GPT-OSS 120B on one H100 and give me the chat link when the server is ready.
  • Show my servers and what they cost per hour; stop the ones I haven't touched since yesterday.

Money and safety

  • The agent cannot spend more than your balance: top-ups happen in the browser only.
  • create_instance, start_instance and destroy_instance are annotated as consequential, so clients ask you before running them.
  • The root password stays out of tool output unless asked for explicitly; your SSH keys are installed on every server.
  • The tools call the gpu.nz REST API with your key: its scopes, the rate limit (120 requests a minute per key) and the account's audit log apply.

По-русски

MCP-сервер gpu.nz подключает аренду видеокарт к ИИ-агентам: Claude, Cursor, Codex и другим. Агент сам найдёт подходящую видеокарту, посчитает видеопамять для модели, арендует сервер с ComfyUI, Ollama или vLLM, прочитает логи и остановит сервер. Адрес: https://gpu.nz/mcp, оплата поминутная, в рублях через СБП или криптой. Документация: gpu.nz/docs/mcp.

License

MIT. gpu.nz is a GPU cloud billed by the minute; the hosted server and the API are run by gpu.nz.

來源：README.md，提交 875bb06

工具

16

list_gpu_models

List GPU models

GPU models available on gpu.nz right now with VRAM, class and the cheapest hourly price per GPU. Use it to pick a gpu_slug for search_offers.

輸入參數querymin_vram_gb

search_offers

Search GPU servers

Search rentable GPU servers (offers) with live prices. Returns the cheapest matches first unless `order` says otherwise. Each result has an offer id for create_instance.

輸入參數gputypelimitordercountrynum_gpusverifiedmin_disk_gbmin_vram_gbmax_price_rub_hmin_reliability

get_offer

Get one offer

Details and the current price of one offer by id (from search_offers).

輸入參數offer_id必填

list_templates

List templates

Ready environments a server can start with: ComfyUI, Stable Diffusion Forge, Ollama + Open WebUI, vLLM, Jupyter/PyTorch, Kohya, FluxGym, Unsloth, chats with preloaded models, plain Ubuntu + CUDA, or a custom Docker image. Pass `slug` as `template` to create_instance.

無需輸入參數。

estimate_vram

Estimate VRAM for an AI model

How much GPU memory an AI model needs (LLMs such as Qwen, DeepSeek, Llama, Gemma, GPT-OSS; image and video models such as FLUX, SDXL, Wan) at a given precision and context length, and which GPUs on gpu.nz fit it with their current price. Call without `model` to list the known models.

輸入參數modelprecisioncontext_tokensconcurrent_requests

get_account

Account and balance

The signed-in account and its balance, hourly burn and runway. Needs a key with user:read and billing:read.

無需輸入參數。

list_instances

List my servers

Your rented servers with status, GPU, price and SSH command. Needs instances:read.

輸入參數status

get_instance

Get one server

Status, SSH command, app URLs (ComfyUI, Jupyter, chat…), price and cost so far of one server. The root password is left out unless include_password is true (your SSH keys are installed on every server).

輸入參數instance_id必填include_password

create_instance

Rent a GPU server

Rents the server of an offer and starts it with a template. Billing starts right away and runs every minute until you stop or destroy it. Confirm the price with the user first. Needs instances:write. The response has the instance id; poll get_instance until status is running (usually 1–5 minutes).

輸入參數nametypeimagedisk_gb必填offer_id必填template必填ssh_key_idsidempotency_key

stop_instance

Stop a server

Stops a running server: the GPU is released and billing drops to the disk only. Data on the disk is kept and the server can be started again (the GPU may then be taken by someone else; gpu.nz finds a replacement when it is). Needs instances:write.

輸入參數instance_id必填

start_instance

Start a server

Starts a stopped server again; GPU billing resumes. Confirm with the user. Needs instances:write.

輸入參數instance_id必填

destroy_instance

Destroy a server

Destroys a server and deletes its disk for good; billing stops completely. Irreversible: confirm with the user. Needs instances:write.

輸入參數instance_id必填

get_instance_logs

Server logs

The last lines of a server's container log: boot progress, app start-up, errors. Needs instances:read.

輸入參數tailinstance_id必填

list_ssh_keys

List SSH keys

SSH public keys on the account; they are installed on new servers. Needs user:read.

無需輸入參數。

add_ssh_key

Add an SSH key

Adds an SSH public key (a line from a .pub file) to the account. Needs user:write.

輸入參數namepublic_key必填

版本歷史

1
  1. v1.0.0最新Oct 19, 2026

評分

已滿足適用檢查項的 100%

該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。

  • 近 30 天內可連通必需25 分
  • 已宣告認證方式必需10 分
  • 已列出工具必需15 分
  • 描述必需15 分
  • 圖示10 分
  • 文件15 分
  • 授權條款10 分