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)
One cold-outreach engine for any target: HOOK/BRIDGE/SOFT-ASK, follow-up ladder, reply playbook.
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。
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": {
"outbound-engine": {
"command": "npx",
"args": [
"-y",
"outbound-engine-mcp"
]
}
}
}在终端中运行这条命令
claude mcp add outbound-engine -- npx -y outbound-engine-mcp粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"outbound-engine": {
"command": "npx",
"args": [
"-y",
"outbound-engine-mcp"
]
}
}
}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": {
"outbound-engine": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"outbound-engine-mcp"
]
}
}
}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。
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6日
注册名称已通过 GitHub 账户 closermethod 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(1.0.2)于4个月前进入注册中心。
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 Node.js。
同一分类中有:Spec Driven Development, Remote Desktop Commander, Symantec Cloud DLP AIops
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.
Governed Symantec Cloud DLP — 14 MCP tools, read-only with audit.
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 cold-outreach method layer for AI agents. One engine, any target: brand deals, UGC, newsletter sponsorships, jobs, podcasts.
Built from the method behind The Outbound Engine.
Disclaimer. Outputs are structured outreach-method frameworks based on documented practice. Results vary by niche, proof quality, and consistency. Not legal advice. Follow platform rules and outreach law for your channel and region.
Every AI agent that drafts cold outreach today has the same failure mode: it writes long, polite, generic pitches that ask a stranger for the biggest possible thing on the first touch. Those messages don't get replies, and outreach that doesn't get replies doesn't get deals.
This MCP encodes one rule and enforces it with tools:
A cold pitch is not trying to close a deal. It is trying to get a reply. The deal closes in the follow-up.
It does NOT send messages and does NOT scrape prospects. It is the method layer: give it a draft, a target type, or a reply — get back structure, violations, and the next move.
| Tool | What it returns |
|---|---|
structure_cold_pitch | HOOK / BRIDGE / SOFT ASK structure + the 4 swap variables (who to pitch, signal to research, proof to lead with, soft-ask object) per target type |
audit_cold_message | PASS/FAIL audit of a draft: money in a cold message, identity-first opener, generic hook, hard ask, length over ~90 words, multiple asks — with a rewrite recipe |
build_followup_ladder | The 5-step ladder (Day 3 / 7 / 12 / 18 / 25) — where 80% of deals actually close. The Day 25 break-up alone pulls ~30% of replies |
handle_reply | The next move for any reply type: price question, interested, not now, no budget, exposure offer, spec request, pay-on-performance, ghosted |
build_pricing_menu | The 3-option menu (Starter / Most Popular / Full) + the two scoping questions to ask first. Drop scope, never price |
detect_red_flags | Scans a reply or brief for walk-away signals: exposure-pay, spec work, pay-on-performance, budget mismatch, scope creep |
get_full_pack | The complete engine as one payload — drop it into a Claude Project system prompt |
// Agent drafted a cold DM for a brand deal. Before sending:
mcp.call("audit_cold_message", { message: "Hi, I'm a UGC creator with 50k followers! I love your brand. My rate is $500/video...", target_type: "brand_deal" }); // Returns: FAIL — no_money_in_cold_message, no_identity_first_opener, // specific_hook_required + a 3-move rewrite recipe.
// Prospect replied asking "what do you charge?" mcp.call("handle_reply", { reply_type: "price_question" }); // Returns: don't send one number — two scoping questions first, then the 3-option menu.
brand_deal · ugc · newsletter_sponsorship · job · podcast · client_work
Same engine, four swapped variables per target. That's the whole trick.
This MCP is the free method layer. The full Outbound Engine product adds per-target research packs, an onboarding intake that builds your proof inventory, and the daily operating system — at elisabethhitz.com.
Elisabeth Hitz — 10+ years of B2B enterprise sales experience across ad-tech, SaaS, media, and global hiring. Now building MCP servers for the AI agent ecosystem.
License: MIT