reelmint-local-mcp
Downloads and analyzes videos from platforms like Instagram, YouTube, and TikTok locally, returning keyframes and optional transcripts.
README
reelmint-local-mcp
A tiny local MCP server for Claude Desktop (and any MCP client) whose one job is to download a video on your own machine — Instagram reels, YouTube, TikTok, or any yt-dlp-supported URL — and hand back the saved file path (plus a few preview keyframes).
Because it runs on your computer (a residential IP, optionally with your logged-in browser cookies), it reaches content a server can't: platforms block datacenter IPs, so a cloud service either fails or has to pay a proxy to fetch the bytes. On your laptop, the download is free and just works.
It does only the download. Understanding the video, transcripts, and building anything from it are done by whatever MCP/tools you hand the file to.
Built for ReelMint but depends on nothing from it — works with any MCP client.
Tools
| Tool | What it does |
|---|---|
download_video(url, frames=6) |
Download a video to a local file → returns the saved .mp4 path + a few keyframe images for a quick look. |
status() |
Reports whether yt-dlp / ffmpeg are installed + the active cookie/save config. |
Requirements
- Python 3.10+
- ffmpeg — for the preview keyframes. macOS:
brew install ffmpeg(downloads still work without it; you just won't get frames)
Install
git clone https://github.com/HiteshR90/reelmint-local-mcp.git
cd reelmint-local-mcp
python3 -m venv .venv
.venv/bin/pip install -U pip -r requirements.txt
# sanity check (expect ready: True)
.venv/bin/python -c "import server; print(server.status())"
Connect to Claude Desktop
Add this to ~/Library/Application Support/Claude/claude_desktop_config.json
(merge into your existing mcpServers), using absolute paths, then fully quit
and reopen Claude Desktop:
{
"mcpServers": {
"reelmint-local": {
"command": "/absolute/path/to/reelmint-local-mcp/.venv/bin/python",
"args": ["/absolute/path/to/reelmint-local-mcp/server.py"],
"env": {
"IG_COOKIES_FROM_BROWSER": "chrome",
"IG_SAVE_DIR": "~/Downloads/reelmint-reels"
}
}
}
}
Then just ask: "download_video this <url>" → you get the saved path (and a preview). Drag that file into your MCP client to do anything else with it.
Configuration (env)
| Var | Default | Meaning |
|---|---|---|
IG_COOKIES_FROM_BROWSER |
chrome |
Browser to read cookies from (chrome/safari/firefox/edge/brave); "" = public only. Unlocks private/followed content. |
IG_SAVE_DIR |
~/Downloads/reelmint-reels |
Where downloads are saved. |
IG_MAX_FRAMES |
8 |
Hard cap on preview keyframes. |
Notes & limits
- Private accounts need your browser cookies (
IG_COOKIES_FROM_BROWSER) and that you follow the account. Reading cookies may prompt for OS keychain access; some browsers must be closed first. - Platforms fight scrapers; keep it fresh with
pip install -U yt-dlp. - This is a client-side utility. You are responsible for complying with the terms of service of the sites you download from.
License
MIT — © 2026 Hitesh R (ReelMint).
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。