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)
Lets a coding agent read the screenshots you beam to its server with Hostbeam.
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。
HOSTBEAM_DIR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": {
"hostbeam": {
"command": "npx",
"args": [
"-y",
"@hostbeam/mcp"
],
"env": {
"HOSTBEAM_DIR": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add hostbeam --env HOSTBEAM_DIR=your-value -- npx -y @hostbeam/mcp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": {
"hostbeam": {
"command": "npx",
"args": [
"-y",
"@hostbeam/mcp"
],
"env": {
"HOSTBEAM_DIR": "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": {
"hostbeam": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@hostbeam/mcp"
],
"env": {
"HOSTBEAM_DIR": "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。
HOSTBEAM_DIR 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
注册名称已通过域名 hostbeam.app 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
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(0.1.4)于29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 Node.js。
同一分类中有:Spec Driven Development, Remote Desktop Commander, Detailing Manager
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.
Connect your auto detailing studio's tasks, cars, clients and revenue to Claude or ChatGPT.
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。
Hostbeam sends a screenshot from your Mac or your iPhone to a server over SSH and puts the remote path on your clipboard, so you can paste it to a coding agent running there.
This MCP server runs on that server, next to the agent. The agent can ask for the screenshot you just beamed and read it, without you pasting the path.
On the server your agent runs on, not on your Mac or your phone. Node 18 or later.
For Claude Code:
claude mcp add -s user hostbeam -- npx -y @hostbeam/mcp
-s user makes it available in every project on that server.
For Codex:
codex mcp add hostbeam -- npx -y @hostbeam/mcp
For other MCP clients, in their JSON config:
{
"mcpServers": { "hostbeam": { "command": "npx", "args": ["-y", "@hostbeam/mcp"] } } }
If the server can't reach npm, copy hostbeam-mcp.mjs to it and run that instead. It has no dependencies.
claude mcp add -s user hostbeam -- node /path/to/hostbeam-mcp.mjs
Hostbeam beams to /tmp/hostbeam unless you changed the folder for that host. If you did, give the server the same folder: -e HOSTBEAM_DIR=/your/folder for Claude Code, --env HOSTBEAM_DIR=/your/folder for Codex, or "env": { "HOSTBEAM_DIR": "/your/folder" } in JSON.
Only files Hostbeam delivered: names starting with hostbeam- with an image extension, in that one folder. It doesn't look in subfolders, and it refuses a name with a path in it. It never writes, moves or deletes anything.
It skips symlinks, and it won't read a folder that belongs to another user or that any user can write to, because a file there may not be one you beamed.
JPEG, PNG, GIF and WebP come back as images, up to 3.75 MB each. Other formats and larger files come back as a path instead: models only accept those four formats, and model APIs refuse bigger images. When a beam holds several files, latest_screenshot includes images up to 7.5 MB in total and names the rest.
It speaks newline-delimited JSON-RPC on stdin and stdout:
printf '%s\n' \
'{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}' \ '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"list_screenshots"}}' \ | npx -y @hostbeam/mcp
MIT