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)
Paid fallback for public webpages: clean text, Markdown, links, metadata, and JavaScript rendering.
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 …”。
适用于网页版、桌面版和手机版 Claude:
https://smartfetch-production-ea53.up.railway.app/mcp在终端中运行这条命令
claude mcp add --transport http smartfetch https://smartfetch-production-ea53.up.railway.app/mcp粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"smartfetch": {
"url": "https://smartfetch-production-ea53.up.railway.app/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": {
"smartfetch": {
"type": "http",
"url": "https://smartfetch-production-ea53.up.railway.app/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。
ChatGPT 只能连接互联网上的服务器:
https://smartfetch-production-ea53.up.railway.app/mcp菜单名称可能因套餐和版本不同而略有差异。
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。
我们还没有读取这个服务器的工具列表;我们会轮流检查各个端点。
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 Friezaaaa 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
尚未检查;我们会轮流检查各个端点
尚未检查;我们会轮流检查各个端点
尚未检查;我们会在检查端点时读取工具列表
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(1.10.6)于2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。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, FOGA node
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.
A Finch OG Agent cell as an MCP server: read the world, dry-run and take ERC-8414 tenders by policy.
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。
This source tree is an emergency containment release. Retrieval is temporarily unavailable. Browser rendering is temporarily unavailable. Every valid POST /fetch request and every invocation of the four existing MCP tools fails before payment verification, DNS, network retrieval, provider work or settlement. There is no payment challenge for an unavailable operation.
V1.11 paid capabilities remain disabled: the eight V1.11 REST variants return ordinary 404 responses, and the two additional MCP tools are absent. The V1.11 Stage 5 benchmark remains offline by default; this release does not authorize a real provider benchmark or activate Exa or Gemini.
This repository version is 1.11.1. It does not claim that a V1.11.1 image, Official MCP Registry entry or x402scan listing has been published. V1.10.6 remains the separately published Official MCP Registry version; publication of this emergency version requires separate review and authorization. Deployment, production configuration, Registry publication and x402scan changes are separate actions.
GET /health remains a free liveness route. Its JSON includes version: "1.11.1", retrieval_available: false, browser_available: false and mode: "emergency_fail_closed"./, /docs, /openapi.json, /llms.txt, /.well-known/x402, /meta, /robots.txt and /sitemap.xml remain free. They identify retrieval as unavailable and advertise zero actionable payable resources.POST /fetch returns finite HTTP 503 retrieval_unavailable. A valid forced-browser request instead returns finite HTTP 503 browser_unavailable. Both include Retry-After: 3600 and Cache-Control: no-store and do not process payment headers.initialize and tools/list remain free. tools/list exposes exactly fetch_webpage, webpage_to_markdown, extract_webpage_text and render_webpage, each explicitly marked unavailable. Calling any of them returns a finite error result before a payment wrapper or retrieval handler.search_and_extract and extract_structured_data are absent from MCP discovery.Successful retrieval responses do not exist in this emergency mode. The additive content_trust: untrusted_external marker is reserved for future successful retrieved results; it is not inserted into emergency errors and does not rewrite source content.
Use an isolated Python environment with the repository's declared dependencies and run the emergency test suites before building. The local container build command is:
docker build -t smartfetch:v1.11.1 .
This is a local image tag, not a remote published V1.11.1 image. The Dockerfile runs as UID/GID 10001:10001 and does not install Chromium, FFmpeg or FFprobe. The built-image gate runs the emergency API/MCP smoke script inside a locally started container. No provider, facilitator, wallet or payment credentials are required for startup or these free checks.
Do not activate V1.11 paid interfaces or restore retrieval while the approved SSRF, cancellation, credential-isolation and cleanup boundaries remain unproven. A security rollback may use only another verified fail-closed image; restoring a retrieval-enabled V1.11.0 image is prohibited. Browser restoration and any retrieval-enabled architecture require separate exact-runtime and final-image containment gates. Publishing to the Official MCP Registry or x402scan requires separate authorization after deployment verification.
Historical V1.10 payment/retrieval details remain available in earlier Git history. They are not instructions to pay or execute retrieval against this emergency service.