Video Watch MCP
Let Claude "watch" videos with you. Send a TikTok, YouTube, or any video link - Claude sees the frames and reads the transcript.
Fully cloud-hosted. No local processing. Works on Claude Desktop, Claude mobile, anywhere MCP works.
Potencialize seu agente
Instalação do publicador. Não é um endpoint do PlugMCP.
Host a copy with our keys and logs.
Ainda não há uma versão empacotada no registro. Siga o README do repositório.
URL remota ou um comando na sua máquina? Como as duas instalações diferem
Rodar no PlugMCP
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
AbrirDo publicador
Let Claude "watch" videos with you. Send a TikTok, YouTube, or any video link - Claude sees the frames and reads the transcript.
Fully cloud-hosted. No local processing. Works on Claude Desktop, Claude mobile, anywhere MCP works.
Let AI agents watch videos — cloud-hosted MCP server with transcript + frame extraction for TikTok, YouTube, and 1000+ platforms
Three tools, pick based on content:
Go to modal.com and sign up. Free tier gives you $30/month in credits - enough for thousands of short videos.
pip install modal
modal token set --token-id YOUR_TOKEN_ID --token-secret YOUR_TOKEN_SECRET
(Get your token from Modal's dashboard after signup)
git clone https://github.com/nekyialabs/video-watch-mcp.git
cd video-watch-mcp
modal deploy mcp_remote.py
You'll get a URL like: https://yourusername--video-watch-mcp-mcp-server.modal.run
Edit your claude_desktop_config.json:
Windows: %APPDATA%\Claude\claude_desktop_config.json
Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
Add under mcpServers:
{
"mcpServers": {
"video-watch": {
"url": "https://yourusername--video-watch-mcp-mcp-server.modal.run"
}
}
}
Restart Claude Desktop. Send any video link and ask Claude to watch it:
"Watch this with me: https://tiktok.com/..."
Claude will see the frames and read the transcript.
Anything yt-dlp supports:
With Modal's free tier ($30/month credits):
You'll never hit the limit with normal use.
You send a link
↓
Claude calls watch_video(url)
↓
Modal spins up a container with ffmpeg + whisper
↓
yt-dlp downloads the video
↓
ffmpeg extracts frames (with timestamps burned in)
↓
Whisper transcribes the audio
↓
Returns frames as images + transcript text
↓
Claude sees everything, you discuss it together
mcp_remote.py - The full MCP server (deploy this)video_watch.py - Standalone video processor with web endpoint (if you just want the API)In mcp_remote.py you can adjust:
fps - Frames per second to extract (default 0.5 = one frame every 2 seconds)max_frames - Maximum frames to return (default 10, max 20)whisper model - Using "base" for speed, can use "small" or "medium" for accuracyPull the latest code and redeploy — that's it:
cd video-watch-mcp
git pull
modal deploy mcp_remote.py
Your MCP URL stays the same, no client config changes needed.
Coming from the old maryfellowes/video-watch-mcp repo? That repo has been removed. Update your remote:
git remote set-url origin https://github.com/nekyialabs/video-watch-mcp.git
git pull
modal deploy mcp_remote.py
MIT - do whatever you want with it.
Built by Codependent AI.
Built by Codependent AI.
Let AI agents watch videos — cloud-hosted MCP server with transcript + frame extraction for TikTok, YouTube, and 1000+ platforms
Ferramentas
Ainda não há lista de ferramentas. Servidores remotos recebem uma após a primeira verificação de saúde.
Perguntas frequentes
Copie nesta página a configuração stdio para Cursor, Claude Desktop, Claude Code ou VS Code. Ela usa a instalação do publicador, não um proxy do PlugMCP.
Este registro não lista nenhuma variável de ambiente obrigatória.
Video Watch roda localmente via stdio.
Última verificação: 2026-10-28. Let AI agents watch videos — cloud-hosted MCP server with transcript + frame extraction for TikTok, YouTube, and 1000+ platforms