MCP YouTube Transcript
An MCP server retrieving transcripts of YouTube videos
作者提供的说明(英文原文)
Official MCP server for WireBoard. Lets LLM agents query your analytics in conversation.
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。
WIREBOARD_TOKEN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 …”。
打开 Claude Desktop → Settings → Developer → Edit Config,把下面的内容粘贴到 claude_desktop_config.json
{
"mcpServers": {
"wireboard": {
"command": "npx",
"args": [
"-y",
"@wireboard/mcp"
],
"env": {
"WIREBOARD_TOKEN": "your-value"
}
}
}
}在终端中运行这条命令
claude mcp add wireboard --env WIREBOARD_TOKEN=your-value -- npx -y @wireboard/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": {
"wireboard": {
"command": "npx",
"args": [
"-y",
"@wireboard/mcp"
],
"env": {
"WIREBOARD_TOKEN": "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": {
"wireboard": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@wireboard/mcp"
],
"env": {
"WIREBOARD_TOKEN": "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。
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 wireboard 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。
你需要填写:WIREBOARD_TOKEN
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(0.1.3)于5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:MCP YouTube Transcript, Google Analytics MCP, BibiGPT YouTube Video Summarizer
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
An MCP server retrieving transcripts of YouTube videos
GA4 MCP server for AI agents: schema discovery, server-side aggregation, analysis-ready defaults.
AI video summarizer: summarize and transcribe YouTube, Bilibili, TikTok and podcast videos.
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。
wireboard-mcp
Official Model Context Protocol server for WireBoard .
Lets LLM agents (Claude Desktop, Cursor, VS Code, etc.) query your real-time and historical analytics in conversation. Built on top of the official @wireboard/api JavaScript SDK.
Two ways to install, pick whichever fits your setup.
Download wireboard-mcp-x.y.z.mcpb from the GitHub releases page and double-click it. One file works on Windows, macOS, and Linux — Claude Desktop ships its own Node runtime, so there are no system dependencies to install.
Claude Desktop will prompt for your WireBoard API token, store it securely in your OS keychain, and the WireBoard tools become available immediately.
If double-click doesn't open the file, install it via Claude Desktop → Settings → Extensions → Advanced Settings → Install Extension.
npm install -g @wireboard/mcp
Requires Node 18+. Then configure your MCP client of choice.
Edit claude_desktop_config.json (Settings → Developer → Edit Config):
{
"mcpServers": { "wireboard": { "command": "wireboard-mcp", "env": { "WIREBOARD_TOKEN": "your_token_here" } } } }
Restart Claude Desktop. The WireBoard tools will appear automatically.
Use the same command + env-var pattern in your client's MCP config.
You need a WireBoard API token before either install path will work. Mint one at Settings → API with the analytics:read ability for REST tools and live:read for the live snapshot tool.
Once configured, ask Claude things like:
Claude will pick the right tool, call it, and answer in natural language.
| Tool | What |
|---|---|
list_sites | Every site in the account |
get_account | Token-owner identity + abilities |
get_aggregate | Period totals: visitors, pageviews, bounce rate, duration |
get_timeseries | One metric (visitors or pageviews) bucketed by hour or day |
get_history | Per-day visitors / returning / pageviews / bounce / duration |
get_breakdown | Top-N rows by dimension (country, device, browser, referrer, etc.) |
get_top_urls | Per-URL metrics with prefix / contains / exact filters |
query_events | Custom event queries with grouping and filtering |
get_live_state | Real-time snapshot (live visitor count, top pages, active sessions, etc.) |
list_dimensions | Meta: every dimension, metric, and limit the API supports |
All tools accept natural date ranges: "today", "yesterday", "last 7 days" (or "30d" shorthand), "this week", "last week", "this month", "last month", or explicit "YYYY-MM-DD..YYYY-MM-DD". Always UTC.
The MCP proactively caps itself at 100 requests/minute (under the API's 120/minute limit) so LLM bursts space themselves out instead of hitting 429s. Override with the WIREBOARD_MCP_RATE_PER_MINUTE env var if you have a use case that needs different pacing.
The underlying SDK still auto-retries on 429 as a backstop.
analytics:read and live:read scope on every site in the account.The MCP is read-only: it can fetch data, never modify it. The WireBoard public API itself is read-only in v1.
Logs go to stderr (so they don't interfere with the MCP protocol on stdout).
@wireboard/mcpgithub.com/wireboard/mcpgithub.com/wireboard/mcp/releasesgithub.com/wireboard/mcp/issues@wireboard/api (source)npm install
npm test # run vitest npm run build # bundle TS → dist/index.js (esbuild, ~600 KB) npm run build:mcpb # also pack dist/wireboard-mcp- .mcpb
The .mcpb is a zip of manifest.json, icon.png, and the single bundled dist/index.js. All runtime dependencies are inlined by esbuild.
MIT.