Runcrate

io.github.aeonmindai由 GitHub 账号 aeonmindai 发布v1.5.1更新于 Nov 3, 2026

面向智能体的AI云：启动GPU、运行可自行停止计费的任务、安排定时运行。

AI 翻译由 AI 模型根据下方的原文描述翻译，可能有误。

原文描述

AI cloud for agents: launch GPUs, run jobs that stop their own billing, schedule runs.

已验证Streamable HTTP可网页运行云端与基础设施AI 与机器学习

概览

Runcrate 的托管 MCP 服务（mcp.runcrate.ai）；需要 OAuth 登录；Streamable HTTP。

暂无 AI 概览：以下信息来自注册中心条目和服务本身。

端点
mcp.runcrate.ai
最近检查

安装

在 SourceWeft 中

  1. 打开 控制台中的 Runcrate，将其添加到工作区。
  2. 连接 所需的 凭据。它们会被私密存储，绝不会在公开页面上展示。
  3. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/io.github.aeonmindai%2Fruncrate/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。 客户端会在首次连接时完成 OAuth 登录。

{
  "mcpServers": {
    "runcrate": {
      "type": "http",
      "url": "https://mcp.runcrate.ai/mcp"
    }
  }
}

README

[Runcrate]

Runcrate MCP

Give your AI agent GPUs. Runcrate is the AI cloud for developers and agents: connect it once, then ask in plain words. Your agent sees live GPU prices, launches a GPU on a ready image, runs your command as a job whose machine stops (and stops billing) when the command exits, schedules runs, moves files, and checks your spend.

  • Server: https://mcp.runcrate.ai/mcp (streamable HTTP)
  • Sign-in: OAuth in your browser. No key to copy, nothing to paste into a config file.
  • Tools for instances, clusters, jobs, schedules, files, volumes, firewall and HTTPS links, billing, environments and secret names. Tool reference

You pick one environment when you sign in. The agent acts only there, and that environment's monthly budget caps what it can spend.

Connect your agent

Paste this one line into your agent: Claude Code, Codex, Cursor, VS Code, OpenCode, or any agent that can run a command. It adds the server to itself, and its first Runcrate tool call opens your browser to sign in.

text
Add Runcrate's MCP server to yourself: run curl -fsSL "https://runcrate.ai/mcp.md" and follow it.

Or one click:

[Add to Cursor] [Install in VS Code] [Install in VS Code Insiders]

In Claude Code, the plugin adds the server and a skill that teaches Claude how to spend safely: /plugin marketplace add runcrateai/runcrate-mcp, then /plugin install runcrate@runcrate. In Gemini CLI: gemini extensions install https://github.com/runcrateai/runcrate-mcp.

The same skill installs into Codex, Cursor, Gemini CLI and other agents that read Agent Skills: npx skills add runcrateai/runcrate-mcp --skill runcrate. It teaches your agent to check live prices (https://www.runcrate.ai/pricing.md), run work with a max runtime, stop what it started, and top up when the balance runs low.

Claude and ChatGPT on the web can't run a command, so they take the URL as a custom connector (steps below).

Set it up by hand

With the Runcrate CLI, one command connects every client it finds on your machine:

bash
runcrate mcp install

Claude Code

bash
claude mcp add --transport http --scope user runcrate https://mcp.runcrate.ai/mcp

Then run /mcp in Claude Code and pick runcrate: your browser opens to sign in.

Claude (claude.ai and Claude Desktop)

  1. Open Settings > Connectors and choose Add custom connector.
  2. Name it Runcrate and paste https://mcp.runcrate.ai/mcp.
  3. Choose Connect and sign in to Runcrate in the window that opens.

ChatGPT

  1. Open chatgpt.com/plugins, choose +, then Add custom MCP server.
  2. Name it Runcrate and paste https://mcp.runcrate.ai/mcp as the URL, with OAuth.
  3. Create it, sign in to Runcrate, then mention @Runcrate in a chat.

Codex

bash
codex mcp add runcrate --url https://mcp.runcrate.ai/mcp

Codex opens your browser to sign in. Closed it? Run codex mcp login runcrate.

Windsurf and Devin

Run runcrate mcp install --client windsurf, or add a server in Cascade's MCP settings with the URL https://mcp.runcrate.ai/mcp. Devin CLI: devin mcp add -s user runcrate https://mcp.runcrate.ai/mcp, then devin mcp login runcrate.

Zed, Warp, Goose, Raycast and any other MCP client

Give the client the URL https://mcp.runcrate.ai/mcp and choose OAuth (dynamic registration). It signs you in through your browser on first use.

Config files

Cursor (~/.cursor/mcp.json), Warp (~/.warp/.mcp.json) and most clients:

json
{ "mcpServers": { "runcrate": { "url": "https://mcp.runcrate.ai/mcp" } } }

VS Code (mcp.json in your user profile):

json
{ "servers": { "runcrate": { "type": "http", "url": "https://mcp.runcrate.ai/mcp" } } }

Windsurf (mcp_config.json): { "mcpServers": { "runcrate": { "serverUrl": "https://mcp.runcrate.ai/mcp" } } }

Zed (settings.json): "context_servers": { "runcrate": { "url": "https://mcp.runcrate.ai/mcp" } }

Gemini CLI (~/.gemini/settings.json): { "mcpServers": { "runcrate": { "httpUrl": "https://mcp.runcrate.ai/mcp" } } }

Codex (~/.codex/config.toml):

toml
[mcp_servers.runcrate]url = "https://mcp.runcrate.ai/mcp"

Goose (~/.config/goose/config.yaml):

yaml
extensions:  runcrate:    name: Runcrate    type: streamable_http    uri: https://mcp.runcrate.ai/mcp    enabled: true

Try it

Launch an H100 with the Finetune crate and run nvidia-smi.

Run python train.py on an A100 as a job, at most 2 hours, and follow the logs.

How much of this month's budget is left?

What is in this repo

FileFor
.claude-plugin/ and .mcp.jsonClaude Code plugin and marketplace
plugin.json and mcp.jsonAgent Plugins package (Cursor, ChatGPT, Codex)
.cursor-plugin/plugin.jsonCursor Marketplace
gemini-extension.jsonGemini CLI extension
skills/runcrate/SKILL.mdThe skill every package ships: how to pick jobs or instances and keep spend in check

The server itself is hosted by Runcrate; this repo only packages the connection. It is listed in the official MCP registry as io.github.aeonmindai/runcrate.

Links

MIT licensed.

来源：README.md，提交 c05ee01

工具

0
该服务器需要登录，因此 SourceWeft 无法在没有账号的情况下读取其工具，也未收录。已连接账号所能看到的内容可能因账号及其权限而异。

版本历史

5
  1. v1.5.1最新Nov 3, 2026
  2. v1.5.0Oct 26, 2026
  3. v1.4.0Oct 25, 2026
  4. v1.1.0Oct 20, 2026
  5. v1.0.0Oct 19, 2026

评分

已满足适用检查项的 85%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list，或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分