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)
Checks if an AI agent may pay: duplicate invoices, totals, evidence. Signed receipts.
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。
Authorization标为机密的值是你账户的钥匙。千万不要分享给别人,也不要粘贴到聊天里。
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 …”。
这个地址需要在请求的 header 中放入密钥,而添加连接器的菜单无法设置。请改用 Claude Code、Cursor 或 VS Code。
在终端中运行这条命令
claude mcp add --transport http screwtape https://screwtape.ai/mcp --header 'Authorization: your-value'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": {
"screwtape": {
"url": "https://screwtape.ai/mcp",
"headers": {
"Authorization": "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": {
"screwtape": {
"type": "http",
"url": "https://screwtape.ai/mcp",
"headers": {
"Authorization": "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。
这个地址需要在请求的 header 中放入密钥,而添加连接器的菜单无法设置。请改用 Claude Code、Cursor 或 VS Code。
你的选择通过 Cookie 保存在这台设备上,无需账户。
Authorization 机密 HTTP header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。
工具名称和描述与服务器在 2026年11月6日 报告的完全一致。这些内容由作者撰写,不是我们写的;我们以纯文本形式显示。
judgeA pass/fail/retry verdict on ONE agent output, from a service outside your agent, used as the go/no-go before an agent spends money or acts on that output. Call it before you pay on, return, store or act on a draft answer, especially when you have the evidence the answer should b
check_duplicate_invoiceSpend gate for invoices: decides whether an agent may pay this invoice, with a signed receipt as proof. Same as judge with one duplicate check: it compares this invoice with the invoices this caller recorded earlier and also judges it. Call it before an agent pays or approves an
consensusFalse-consensus check for 2 to 16 answers to the same task, so a group of agents does not approve a spend just because they copied each other. It scores how the answers agree and flags agreement that looks copied or too thin to trust. Screwtape is not tied to the agent framework
get_api_keyCreate a free Screwtape.ai API key (10 calls per UTC day) with no human approval. Call it only when no Screwtape key is configured and you can compute SHA-256 with a code tool, because the proof of work must be computed exactly, not guessed. Do NOT call it when a key already exis
publish_receiptPublish a signed receipt you own so a public page and a Screwtape Verified badge can link to that one verdict. Call it when you want to show that a specific output received a specific verdict. Do NOT call it to certify an agent in general: the page says this output received this
try_checkSee what a spend check returns before you have a key. Runs ONE sample check on a fixed demo invoice: a duplicate match against an invoice paid earlier, required fields, and a sum check that the line items add up to the total. Call it once to learn the response shape. Do NOT call
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
注册名称已通过域名 screwtape.ai 验证(注册中心会核实域名所有权)
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
在 310 毫秒内响应(HTTP 200)
无需登录
已读取 6 个工具描述,未发现针对 AI 的隐藏指令
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 3 项适用的基础检查中通过了 3 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。
你需要填写:Authorization
它在作者的云端运行。通过互联网连接,无需安装,但你发送的内容会经过作者的服务器。
GitHub 仓库最后更新于–,最新版本(1.2.1)于3天前进入注册中心。
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,所以接受远程 MCP 服务器的应用也可以使用,例如开启开发者模式的 ChatGPT。
同一分类中有:Spec Driven Development, Remote Desktop Commander, zanadir
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.
Find what a repository's CI is missing, and which security checks cannot fail the build.
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。
Copy-paste examples that stop an AI agent from paying the wrong invoice. Each one calls the public Screwtape.ai API with an API key and does two things before money moves:
The agent pays only when next_step.action is pass. Anything else (retry or escalate) means do not pay.
| Folder | Stack | Run |
|---|---|---|
curl/ | curl + jq | ./curl/check_payment.sh |
python/ | Python 3.9+, standard library only | python python/check_payment.py |
typescript/ | Node 18+, built-in fetch | cd typescript && npm i && npm start |
langchain/ | LangChain tool | python langchain/spend_tool.py |
crewai/ | CrewAI tool | python crewai/spend_tool.py |
openai-agents/ | OpenAI Agents SDK, guarded payment tool | python openai-agents/spend_guard.py |
export SCREWTAPE_API_KEY=... # free key: https://screwtape.ai/signup/
python python/check_payment.py
Expected output (receipt ids differ per run):
first: verdict=pass action=pass reason=none duplicate=unique receipt=...
-> PAY second: verdict=... action=escalate reason=duplicate_invoice duplicate=duplicate receipt=... -> DO NOT PAY wrong total: verdict=fail action=... reason=sum_mismatch duplicate=... receipt=... -> DO NOT PAY
result = check_payment(invoice, evidence) # POST https://screwtape.ai/v1/duplicate
if result"next_step" != "pass": stop_and_escalate(result) # do not pay
Put the check inside the payment tool itself (see openai-agents/), so the model cannot skip it.
POST /v1/duplicate: spend check plus duplicate-invoice check (used by every example here).POST /v1/judge: verdict on any agent output, with deterministic checks (sum, required_fields, json_schema, regex, equals, ...). See checks.POST /v1/consensus: compare 2 to 16 agent answers before acting on them.POST /v1/receipts/verify: verify a signed receipt.Authorization: Bearer $SCREWTAPE_API_KEY. Add an Idempotency-Key header to make retries safe.Agents that speak MCP can connect directly: https://screwtape.ai/mcp (Streamable HTTP, Bearer key). Tools: judge, check_duplicate_invoice, consensus.
Details: screwtape.ai/pricing.
MIT