youtube-mcp-tools
Evidence-first YouTube research server enabling search, transcript citation, video analysis, flashcards, and collateral generation for MCP hosts and VS Code.
README
YouTube MCP Tools
Evidence-first YouTube research for modern MCP hosts and VS Code.
The project consolidates the supported YouTube MCP and extension workflows:
- search YouTube and inspect videos, channels, and playlists;
- retrieve, search, and cite caption transcripts;
- analyze video content and create flashcards;
- research a topic into a transparent, curated video collection;
- turn selected, transcript-backed videos into host-grounded or direct-provider collateral;
- render accessible static HTML resource pages.
Evidence-first collateral
The research workflow does not treat popularity or a hidden score as proof of quality. Each selected video carries explicit evidence for topic relevance, requested coverage, transcript availability, source identity, practical duration, and optional recency. The output includes selections, exclusions, and a verification date.
Collateral supports two intentional modes:
- Host-managed generation returns a source-constrained prompt for Copilot, Claude, or another MCP host to write original articles, summaries, or study guides.
- Direct-provider generation calls an explicitly configured provider through a small adapter interface. No provider credential or network integration is embedded in the core.
Both modes require timestamp citations to selected source videos and instruct the writer to paraphrase rather than reproduce transcript language.
Reference integration
defensibleDecisionDashboardBrief defines the first editorial profile: a
decision-ready Power BI dashboard guide for The Defensible Decision. It will produce
a curated video page and a companion study guide for the book's /resources/ section.
The reader-facing guide is published at Power BI Video Guide: From First Report to Defensible Decision. It pairs selected Power BI tutorials with the book's question-first sequence for building and reviewing a report.
VS Code extension
The maintained Marketplace extension source is in
apps/vscode-extension. It retains the published
fabioc-aloha.youtube-mcp-tools extension identifier and its existing search,
analysis, transcript, flashcard, quota, and Activity Bar workflows. Its bundled
MCP executable is built from this canonical server, so extension updates carry
the same modern host integration as the npm distribution.
Development
npm install
npm test
Install in MCP hosts
Local build
The 2.0 source release is available now. Install dependencies and build it:
npm install
npm run build
Then add this server entry to an MCP host:
{
"servers": {
"youtube": {
"type": "stdio",
"command": "node",
"args": ["C:\\absolute\\path\\to\\youtube-mcp-tools\\dist\\mcp-server\\index.js"]
}
}
}
This is compatible with VS Code's workspace .vscode/mcp.json and user
configuration. Transcript tools work with no configuration. To use search,
metadata, analysis, flashcards, or live topic research, add YOUTUBE_API_KEY
through your host's secret or environment-variable facility; never commit a
key.
npm package
The npm publication of youtube-mcp-tools@2.0.1 is pending. After it is
published, replace the local node command above with:
{
"type": "stdio",
"command": "npx",
"args": ["-y", "youtube-mcp-tools@2.0.1"]
}
Get a YouTube Data API key
Transcript retrieval and transcript search do not require a Google API key. You need a key for YouTube search, video metadata, analysis, flashcards, and live topic research.
- In Google Cloud Console, create or select a project.
- Open APIs & Services → Library, search for YouTube Data API v3, and enable it for that project.
- Open APIs & Services → Credentials → Create credentials → API key.
- Restrict the key's API restrictions to YouTube Data API v3. If the server runs from a stable egress address, also use an IP-address application restriction. Do not use a browser-referrer restriction for this local stdio server.
- Store the key only in your MCP host's secret or environment-variable
facility. For local development, copy
.env.exampleto.env;.envis ignored by Git. The server loads it only whenYOUTUBE_API_KEYis not already defined in the environment.
Never paste a key into an issue, source file, committed configuration, or a prompt. If a key is exposed, delete or rotate it in Google Cloud Console.
For Claude Code, add the local build with:
claude mcp add youtube --scope project --env YOUTUBE_API_KEY=your_key -- node dist/mcp-server/index.js
The server also exposes the youtube_research_to_collateral MCP prompt for
hosts that support prompts.
GitHub Copilot CLI through Alex ACT Plugin Mall
After the npm package is published, GitHub Copilot CLI users can install the optional Mall plugin:
copilot plugin marketplace add fabioc-aloha/Alex_Skill_Mall
copilot plugin install youtube-research@alex-mall
The plugin declares the same youtube stdio server and uses the exact npm
release. This is a Copilot-specific convenience route; Claude, Cursor, VS
Code, and other MCP hosts should use the standard installation above.
Marketplace readiness
server.json is the official MCP Registry manifest. After publishing the npm
package, publish this manifest with the official
MCP Registry publisher.
The package contains the matching mcpName ownership marker required by the
registry.
The manual GitHub Actions workflow at
.github/workflows/publish-release.yml publishes a pushed v<version> tag in
order: npm package, then MCP Registry manifest. Configure either npm trusted
publishing or an NPM_TOKEN repository secret; the MCP Registry step uses
GitHub OIDC and needs no dedicated secret. Publish the VS Code companion VSIX
separately through the Visual Studio Marketplace publisher portal.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。