MCP YouTube Transcript
An MCP server retrieving transcripts of YouTube videos
作者提供的说明(英文原文)
Transcript retrieval for YouTube, TikTok and Instagram: whole channels, playlists and searches.
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://transcriptfetch.com/mcp在终端中运行这条命令
claude mcp add --transport http youtube-transcripts https://transcriptfetch.com/mcp粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"youtube-transcripts": {
"url": "https://transcriptfetch.com/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": {
"youtube-transcripts": {
"type": "http",
"url": "https://transcriptfetch.com/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://transcriptfetch.com/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
注册名称已通过域名 transcriptfetch.com 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
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(1.2.3)于14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: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。
A Model Context Protocol server that gives any MCP client (Claude Desktop, Cursor, and others) access to the TranscriptFetch API: a production transcript API for YouTube, TikTok and Instagram, with AI transcription when captions are missing. Fetch transcripts, search videos, list channels and playlists, and check your credit balance.
tf_live_.TRANSCRIPTFETCH_API_KEY in the client configuration below.Every account gets 50 free credits a month. Failures are free: a fetch that returns no transcript is never charged.
Runs locally over stdio and calls the TranscriptFetch API with your key. Prefer a hosted, remote server? Point your client at https://transcriptfetch.com/mcp instead (OAuth or API key). The hosted server waits inline for short-form AI transcription, so no polling is needed there.
| Tool | What it does |
|---|---|
get_transcript | Transcript for a video: YouTube, TikTok, Instagram, or a direct media URL. When a video has no captions but has speech, the audio is transcribed automatically; ai_fallback: true goes straight to the audio. Optional markdown (timestamped, linked paragraphs), chunks (about 500 tokens each, for a vector store) and timestamps: false for text only |
search_videos | Search YouTube (default), TikTok or Instagram. On YouTube, filter by upload_date, duration and captions, and sort by views |
search_channels | Find YouTube channels (default) or playlists by keyword |
list_channel_videos | A YouTube channel's videos, Shorts or live streams (tab), sorted newest, popular or oldest, or searched with query; also a TikTok or Instagram profile |
list_channel_playlists | A YouTube channel's playlists, or its podcasts |
list_playlist_videos | A YouTube or TikTok playlist's videos |
get_credits | Remaining credit balance for the key. Never billed |
These are the same seven tools, with the same inputs, as the hosted server at https://transcriptfetch.com/mcp: npm run sync-tools regenerates them from its public card.
Pricing is per successful result: a caption fetch or a video list costs 1 credit, and AI transcription of the audio costs 1 credit per started 5 minutes of audio, charged only on delivery. Failed, blocked and empty results are never charged, which matters on short-form video where many clips have no speech at all.
No install needed. Run it on demand with npx:
TRANSCRIPTFETCH_API_KEY=tf_live_... npx -y transcriptfetch-mcp
Or install globally:
npm install -g transcriptfetch-mcp
Requires Node 18+.
git clone https://github.com/TranscriptFetch/mcp-server
cd mcp-server && npm install && npm run build
Then point your client at the built entrypoint with "command": "node" and "args": ["/absolute/path/to/mcp-server/dist/index.js"].
Add this to claude_desktop_config.json (Settings then Developer then Edit Config):
{
"mcpServers": { "transcriptfetch": { "command": "npx", "args": ["-y", "transcriptfetch-mcp"], "env": { "TRANSCRIPTFETCH_API_KEY": "tf_live_..." } } } }
Add the same block under mcpServers in your Cursor MCP settings.
Restart the client, and the five tools appear.
Once connected, ask your assistant naturally:
Get the transcript for https://youtu.be/aircAruvnKk and summarize the key points.
Search YouTube for "how transformers work" and list the top 5 videos.
List the latest videos from @lexfridman and pull the transcript of the newest one.
How many TranscriptFetch credits do I have left?
The assistant picks the matching tool and works from the returned transcript or video list.
| Env var | Required | Default |
|---|---|---|
TRANSCRIPTFETCH_API_KEY | yes | none |
TRANSCRIPTFETCH_BASE_URL | no | https://transcriptfetch.com |
The server speaks MCP over stdio, so there is no port to expose. -i is required: without an attached stdin the transport closes immediately and the container looks like it crashed.
docker build -t transcriptfetch-mcp .
docker run --rm -i -e TRANSCRIPTFETCH_API_KEY=tf_live_... transcriptfetch-mcp
MIT