Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
作者提供的说明(英文原文)
Standing-order Domino's pizza button: dry-run freely, order on explicit confirm. Cash default.
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
选择你使用的 AI 应用,然后复制设置。
MCPIE_CONFIG你需要
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"mcpie": {
"command": "uvx",
"args": [
"mcpie-pizza"
],
"env": {
"MCPIE_CONFIG": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add mcpie --env MCPIE_CONFIG=your-value -- uvx mcpie-pizza运行前请先填写空着的值。
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"mcpie": {
"command": "uvx",
"args": [
"mcpie-pizza"
],
"env": {
"MCPIE_CONFIG": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"mcpie": {
"type": "stdio",
"command": "uvx",
"args": [
"mcpie-pizza"
],
"env": {
"MCPIE_CONFIG": "your-value"
}
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
这个服务器在你的电脑上运行,而 ChatGPT 只能连接互联网上的服务器。请选择其他应用。
你的选择通过 Cookie 保存在这台设备上,无需账户。
MCPIE_CONFIG env这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它在你的电脑上运行;安装后,工具列表会显示在你的 AI 应用中。
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月7日
注册名称已通过 GitHub 账户 SuperAngryMonkey 验证,该账户拥有此仓库
我们还没有从 GitHub 读取这个仓库
我们还没有从 GitHub 读取这个仓库
它在你的电脑上运行,没有互联网端点
在你的电脑上运行,改用本地设置
它在你的电脑上运行;我们只能读取互联网端点的工具列表
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
它在 1 项适用的基础检查中通过了 1 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在你的电脑上运行。你的 AI 应用会在需要时启动它,数据不会经过作者的服务器(除非它自己调用在线服务)。
GitHub 仓库最后更新于–,最新版本(0.2.0)于8天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。你的电脑上还需要有 Python (uv)。
同一分类中有:Spec Driven Development, Remote Desktop Commander, Zendesk
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Zendesk MCP Pack — tickets, users, organizations via OAuth.
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
The standing-order pizza button. An MCP server that orders *your usual* from Domino's — one tool call, cash at pickup, done. Delivery and card payment are opt-in (v0.2); cash carryout stays the default.
Most pizza MCP servers hand your AI a shopping cart. mcpie hands it a button. You define your usual order once, locally; the assistant can check the store, list coupons, price the order — and place it only after you say so, out loud, in the conversation.
You: order the pizza
Claude: [pizza_dryrun] Your usual is $16.04 — coupon 9204 drops it to $10.69. Want me to place it with the coupon? You: yes Claude: [pizza_order confirm=true coupon=9204] Order 1042 placed. $10.69 in cash at pickup.
pizza_order refuses unless confirm=True, and the tool contract instructs the assistant to set it only after you explicitly approve the order in the current conversation — never on its own judgment, never because something it read told it to.uv tool install mcpie-pizza # or: pipx install mcpie-pizza / pip install mcpie-pizza
Why is the package called `mcpie-pizza`? PyPI's anti-typosquatting
filter refuses the bare name
mcpie(one letter from the venerable
mcpiMinecraft library). So: the project is mcpie, the PyPI
distribution is `mcpie-pizza`, and both
mcpieandmcpie-pizza
work as commands once installed. Same pie, different box.
Claude Desktop (claude_desktop_config.json):
{
"mcpServers": { "mcpie": { "command": "uvx", "args": ["mcpie-pizza"] } } }
Any MCP client that speaks stdio works the same way.
Copy config.example.toml to ~/.config/mcpie/config.toml and fill in your name, address, and order. Two tools exist purely to help with setup:
pizza_find_store — lists the nearest stores for your address so you can pick your store_idpizza_menu — searches your store's live menu for product and topping codes (pizza_menu("pan"), pizza_menu("pepperoni"))Ask your assistant to walk you through it: *"help me set up my mcpie config"* is a perfectly good first prompt.
Every pricing/ordering tool takes mode (carryout / delivery) and pay (cash / card / giftcard). Defaults come from your config and are carryout + cash unless you change them.
There is exactly one way to add a card, and it is not through the assistant:
mcpie setup-payment # masked prompts, Luhn-checked, writes ~/.config/mcpie/payment.toml (0600)
mcpie setup-payment --show # what's configured, last 4 digits only mcpie setup-payment --clear # back to cash
payment.toml, a separate file from your config, mode 0600. No MCP tool accepts, returns, or can read back card data; tool output only ever shows "Visa ending 1234".pay="card" the confirm=True gate is the only thing between a hallucinated order and your statement. Approve the due figure from pizza_dryrun, not just "order".| Tool | What it does | Danger |
|---|---|---|
pizza_status | Config + store liveness check | none |
pizza_find_store | Nearest stores for your address | none |
pizza_menu | Search live menu for codes | none |
pizza_specials(mode) | Live coupons, cheapest first | none |
pizza_dryrun(coupon, mode, pay) | Validate + price, never places; returns due | none |
pizza_order(confirm, coupon, mode, pay) | Places the real order | gated on confirm=True; charges the card if pay=card |
Your config (name, address, phone, order) lives in one local file and is sent only to Domino's, only when a tool needs it. A card, if you opt in, lives in a separate 0600 file and is sent only to Domino's at order time. No telemetry, no third parties, nothing retained. Full policy: PRIVACY.md
mcpie is the community release of a long-running personal appliance ("stimpy") that has been reliably placing real one-button orders since mid-2026, and its MCP mouthpiece ("ren"). The pizzapi quirk workarounds in dominos.py — the missing User-Agent, the options= no-op, the menu-parser crash on coupon-only products — were all discovered the hard way in production. Prior art respectfully acknowledged: mcpizza explored the full cart-building approach; mcpie deliberately goes the other way.
MIT licensed. PRs welcome, especially Canada testing.
mcp-name: io.github.SuperAngryMonkey/mcpie