YouTube MCP

YouTube MCP

Provides production-grade tools for YouTube channel resolution, video metadata extraction, transcripts, and playlist management. It features a quota-aware, AI-friendly design that supports structured searching and listing of public YouTube data.

Category
访问服务器

README

YouTube MCP (youtube_mcp)

PyPI version Python 3.13+ License: MIT

A Model Context Protocol (MCP) server that provides production-grade YouTube inventory tools (channel resolution, channel videos, channel playlists, playlist videos) designed to be safe-by-default, quota-aware, and AI-friendly.

Repository: https://github.com/glonorce/youtube_mcp PyPI: https://pypi.org/project/glonorce-youtube-mcp/

Naming note:

  • Repo name: youtube_mcp
  • Python package / import name: youtube_mcp
  • MCP entry module: python -m youtube_mcp.server

Prerequisites (PATH)

Make sure these commands work in your terminal (they must be on PATH):

  • python --version
  • python -m pip --version
  • pipx --version (only if you use the pipx-based MCP config)

<details> <summary>✅ What is <code>PATH</code>?</summary>

PATH is the list of folders that your OS searches when you run a command like python or pipx.

  • If a command is not on PATH, you'll see errors like:
    • Windows: executable file not found in %PATH%
    • PowerShell: ... is not recognized as the name of a cmdlet

Quick checks (Windows):

where python
where pipx
$env:Path

</details>

On Windows:

  • install Python from https://www.python.org/downloads/
  • during installation, check "Add Python to PATH"

Install pipx (only if you use the pipx-based MCP config):

python -m pip install --user pipx
python -m pipx ensurepath

Notes:

  • After pipx ensurepath, restart your terminal/IDE so pipx is available on PATH.
  • Verify with: pipx --version

✨ Available on PyPI

pip install glonorce-youtube-mcp

# run
youtube_mcp

Screenshots

<!-- GitHub README does not allow JavaScript, so we can't implement a true clickable carousel. This is a "scroll-snap" horizontal gallery (swipe/trackpad/shift+wheel to move sideways). -->

<div style="display:flex; gap:12px; overflow-x:auto; padding: 8px 0; scroll-snap-type:x mandatory; -webkit-overflow-scrolling: touch;"> <a id="screenshot-test-plan" style="scroll-snap-align:start; flex:0 0 auto;" href="#screenshot-test-plan"> <img src="screenshots/test_plan.png" alt="Test plan" style="height: 360px; display:block;" /> </a> <a id="screenshot-test-execution" style="scroll-snap-align:start; flex:0 0 auto;" href="#screenshot-test-execution"> <img src="screenshots/test_execution.png" alt="Test execution" style="height: 360px; display:block;" /> </a> <a id="screenshot-success-result" style="scroll-snap-align:start; flex:0 0 auto;" href="#screenshot-success-result"> <img src="screenshots/success_result.png" alt="Success result" style="height: 360px; display:block;" /> </a> </div>


Features

  • 7 MCP tools (1 text-output tool + 6 structured-output tools)
  • Safe defaults (public data only; shorts/live excluded unless enabled)
  • Quota-aware design (budgeting + expensive strategies as explicit opt-in)
  • Tool discovery works in MCP hosts (tools registered with rich docstrings)
  • Resilient YouTube Data API client (endpoint allowlist, retries, gzip handling)

Tools

This server exposes 7 tools:

  1. get_yt_video_info(video_id_or_url) -> str

    • Accepts a raw video id or a full YouTube URL.
    • Returns video metadata + transcript (when available).
  2. resolve_youtube_channel(channel_ref, ...) -> dict

  3. list_youtube_channel_videos(channel_ref, ...) -> dict

    • Paginated.
    • Returns nextPageToken and next_page_token (alias).
  4. list_youtube_channel_playlists(channel_ref, ...) -> dict

    • Paginated.
    • Returns nextPageToken and next_page_token (alias).
  5. list_youtube_playlist_videos(playlist_id, ...) -> dict

    • Paginated.
    • Returns nextPageToken and next_page_token (alias).
  6. search_youtube_channel_videos(channel_ref, query, ...) -> dict

    • Keyword search within a channel (quota-expensive, uses Search API).
    • Paginated.
  7. list_youtube_video_comments(video_id, ...) -> dict

    • Public comment threads for a video (comments may be disabled).
    • Paginated.

All tools except get_yt_video_info return structured JSON and are documented with AI-friendly docstrings.


Installation & MCP configuration

PyPI distribution name: glonorce-youtube-mcp

Repo name: youtube_mcp · import name: youtube_mcp

API key

All inventory/search/comment tools require a YouTube Data API v3 key:

  • YOUTUBE_API_KEY

Option 1 (recommended for MCP hosts): pipx (no repo clone)

Use this when you want a "zero local project" setup (no clone, no venv management).

{
  "mcpServers": {
    "youtube_mcp": {
      "command": "pipx",
      "args": ["run", "--spec", "glonorce-youtube-mcp", "youtube_mcp"],
      "env": {
        "YOUTUBE_API_KEY": "YOUR_KEY"
      }
    }
  }
}

Option 2: pip into a dedicated venv (no repo clone)

python -m venv .venv_youtube_mcp
# Windows: .venv_youtube_mcp\Scripts\activate
# macOS/Linux: source .venv_youtube_mcp/bin/activate

python -m pip install -U pip
python -m pip install glonorce-youtube-mcp

MCP host JSON example (simple):

{
  "mcpServers": {
    "youtube_mcp": {
      "command": "youtube_mcp",
      "env": {
        "YOUTUBE_API_KEY": "YOUR_KEY"
      }
    }
  }
}

If your MCP host can't find youtube_mcp on PATH (Windows error like: executable file not found in %PATH%), use an absolute path.

MCP host JSON example (absolute path to venv Python + module):

{
  "mcpServers": {
    "youtube_mcp": {
      "command": "C:\\PATH\\.venv_youtube_mcp\\Scripts\\python.exe",
      "args": ["-m", "youtube_mcp.server"],
      "env": {
        "YOUTUBE_API_KEY": "YOUR_KEY"
      }
    }
  }
}

Option 3: local clone (development)

git clone https://github.com/glonorce/youtube_mcp.git
cd youtube_mcp

python -m venv .venv
# Windows: .venv\Scripts\activate
# macOS/Linux: source .venv/bin/activate

python -m pip install -e .

MCP host JSON example (local clone):

{
  "mcpServers": {
    "youtube_mcp": {
      "command": "C:\\PATH\\youtube_mcp\\.venv\\Scripts\\python.exe",
      "args": ["-m", "youtube_mcp.server"],
      "env": {
        "YOUTUBE_API_KEY": "YOUR_KEY"
      }
    }
  }
}

Getting Your YouTube API Key

While not required, you can optionally set up a YouTube Data API key for enhanced functionality. Here's how to get one:

Step 1: Create a Google Cloud Project

  1. Go to the Google Cloud Console
  2. Click "Select a project" at the top of the page
  3. Click "New Project" and give it a name (e.g., "MCP YouTube Extract")
  4. Click "Create"

Step 2: Enable the YouTube Data API

  1. In your new project, go to the API Library
  2. Search for "YouTube Data API v3"
  3. Click on it and then click "Enable"

Step 3: Create API Credentials

  1. Go to the Credentials page
  2. Click "Create Credentials" and select "API Key"
  3. Your new API key will be displayed - copy it immediately
  4. Click "Restrict Key" to secure it (recommended)

Step 4: Restrict Your API Key (Recommended)

  1. In the API key settings, click "Restrict Key"
  2. Under "API restrictions", select "Restrict key"
  3. Choose "YouTube Data API v3" from the dropdown
  4. Click "Save"

Step 5: Set Up Billing (May be required)

  1. Go to the Billing page
  2. Link a billing account to your project
  3. Note: YouTube Data API has a free tier of ~10,000 units per day, which is typically sufficient for most use cases

API Key Usage Limits

  • Free Tier: ~10,000 units per day (commonly)
  • Cost: depends on your Google Cloud / billing configuration

Security Best Practices

  • Never commit your API key to version control
  • Use environment variables as shown in the MCP configuration section
  • Restrict your API key to only the YouTube Data API
  • Monitor usage in the Google Cloud Console

Example MCP calls (JSON payloads)

Exact envelope depends on your MCP client; these show tool + arguments.

Resolve a channel

{
  "tool": "resolve_youtube_channel",
  "arguments": {
    "channel_ref": "@GoogleDevelopers",
    "resolution_mode": "strict",
    "include_uploads_playlist": true
  }
}

List channel videos (safe default)

{
  "tool": "list_youtube_channel_videos",
  "arguments": {
    "channel_ref": "@GoogleDevelopers",
    "max_videos": 50,
    "page_token": null,
    "include_shorts": false,
    "include_live": false,
    "parts_level": "basic",
    "order_strategy": "uploads_playlist",
    "order_by": "date"
  }
}

Quota reference (best-effort)

This section is a best-effort summary and may change over time. Always verify with official documentation.

Last verified: 2026-02-18

Official references:

  • YouTube Data API quota calculator / costs: https://developers.google.com/youtube/v3/determine_quota_cost
  • Usage limits: https://developers.google.com/youtube/v3/getting-started#quota

Approximate quota usage by tool

Note: get_yt_video_info returns text (not structured JSON) and is not included in the quota table because its behavior depends on optional extractors.

Tool Primary endpoints Typical cost (rough) Notes
resolve_youtube_channel channels.list ~1 strict mode mostly channels.list; best_effort may use search.list
list_youtube_channel_videos (default) channels.list + playlistItems.list + videos.list ~ (1 + 1 + 1) per page uploads-playlist strategy
list_youtube_channel_videos (order_strategy=search_api) search.list + videos.list ~ (100 + 1) per page expensive; capped behavior
list_youtube_channel_playlists playlists.list ~1 per page public playlists only
list_youtube_playlist_videos playlistItems.list + videos.list ~ (1 + 1) per page public playlist
search_youtube_channel_videos search.list + videos.list ~ (100 + 1) per page keyword search within a channel
list_youtube_video_comments commentThreads.list ~1 per page comments may be disabled (403)

Daily quota:

  • Common free quota: 10,000 units/day
  • Reset: daily (Google-defined; see official docs/console)

Testing

Use the venv interpreter to avoid accidentally running global pytest.

# recommended
python -m pytest -q

# if you use a venv
.venv\Scripts\python.exe -m pytest -q

Troubleshooting

Only get_yt_video_info is visible (other tools missing)

Your MCP host is running an old installed copy from site-packages. Fix:

python -m pip install -e .

Then restart the MCP host.

YOUTUBE_API_KEY placeholder is not expanded in some MCP hosts

Some MCP hosts do not interpolate placeholders like ${YOUTUBE_API_KEY} inside mcp.json.

❌ Not recommended (may be taken literally):

{
  "env": { "YOUTUBE_API_KEY": "${YOUTUBE_API_KEY}" }
}

✅ Recommended (explicit value in config):

{
  "env": { "YOUTUBE_API_KEY": "YOUR_REAL_KEY_HERE" }
}

✅ Also works (set it in the host process environment, then reference nothing in JSON):

  • set YOUTUBE_API_KEY in the MCP host environment (OS/IDE)
  • then keep mcp.json without placeholders

License

MIT — see LICENSE


Acknowledgements

This project was originally derived from / inspired by https://github.com/sinjab/mcp_youtube_extract. The codebase has since been heavily rewritten and extended.

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
Kagi MCP Server

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。

官方
精选
Python
graphlit-mcp-server

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。

官方
精选
TypeScript
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选