What Platforms Does Claude Video Support and How to Install Them

Claude Video supports Claude Code, Codex/Cursor/Copilot/Gemini CLI, the claude.ai web interface, and manual developer installations, using a single self-contained skill folder that works across all Agent Skills-compatible hosts.

Claude Video is an open-source Agent Skills package developed by bradautomates that enables video analysis capabilities across multiple AI coding environments. Because it implements the universal Agent Skills specification, the same core code in skills/watch/ runs identically on every supported platform without host-specific modifications.

Supported Platforms Overview

Claude Video operates as an Agent Skills package, making it compatible with any host that supports the Agent Skills protocol. The official installation methods cover four primary platform categories.

Claude Code (Native IDE)

Claude Code is the recommended and primary supported platform. As the native Claude IDE, it provides the smoothest integration for the video analysis workflow.

Install via the built-in plugin marketplace:

/plugin marketplace add bradautomates/claude-video
/plugin install watch@claude-video

Updates are handled automatically through /plugin update watch@claude-video.

Codex, Cursor, Copilot, and Gemini CLI

Over 50 AI coding assistants including Codex, Cursor, GitHub Copilot, and Gemini CLI support Claude Video through the Node-based skills CLI.

Install globally for system-wide access:

npx skills add bradautomates/claude-video -g

Install for the current project only (omitting the -g flag):

npx skills add bradautomates/claude-video

Target specific hosts explicitly:

npx skills add bradautomates/claude-video -a codex -a cursor

The npx skills command discovers the SKILL.md file in the repository and copies the skills/watch folder to the appropriate user skill directory (e.g., ~/.codex/skills or ~/.cursor/skills). Available flags include -l to list skills, -g for global installation, and --copy to force a file copy instead of a symlink.

claude.ai Web Interface

The claude.ai web platform supports Claude Video through manual skill bundle uploads.

1. Download the watch.skill bundle from the latest GitHub release
2. Navigate to Settings → Capabilities → Skills
3. Click the "+" button and upload the watch.skill file

Ensure you enable "Code execution and file creation" under Capabilities, as the skill requires executing ffmpeg and yt-dlp commands.

Manual Developer Installation

For development or custom environments, install Claude Video manually by cloning the repository and symlinking the skill folder.

Clone the repository:

git clone https://github.com/bradautomates/claude-video.git

Create a symlink to your host's skill directory:

ln -s "$(pwd)/claude-video/skills/watch" ~/.codex/skills/watch

Or copy the folder instead of symlinking:

cp -r claude-video/skills/watch ~/.claude/skills/watch

Cross-Platform Architecture

Claude Video achieves portability through a self-contained skill structure defined in skills/watch/SKILL.md. This contract file specifies the /watch slash command and resolves script paths relative to the skill directory.

Key components include:

• skills/watch/SKILL.md — The canonical skill contract defining the /watch command and entry points
• skills/watch/scripts/watch.py — The main orchestrator handling download, frame extraction, and transcription
• skills/watch/scripts/setup.py — Pre-flight dependency checker for ffmpeg, yt-dlp, and Whisper API configuration
• skills/watch/scripts/download.py, frames.py, transcribe.py, whisper.py — Modular utilities for video processing

Because all paths resolve relative to the skill directory, the same Python scripts execute identically whether called from Claude Code, Cursor, or the web interface.

Building the Skill Bundle

Developers can build the .skill bundle manually for the web interface or distribution:

bash skills/watch/scripts/build-skill.sh

This produces dist/watch.skill, a single-file bundle containing the entire skills/watch directory structure.

Verifying Your Installation

After installation, verify Claude Video is recognized by your host:

1. Open your AI assistant (Claude Code, Cursor, etc.)
2. Type /watch followed by a video URL or local file path
3. The setup.py script will automatically check for ffmpeg and yt-dlp dependencies on first run

If dependencies are missing, the skill prompts you to install them before proceeding with video analysis.

Summary

• Claude Code users should use /plugin marketplace add and /plugin install for automatic updates
• Codex/Cursor/Copilot/Gemini CLI users should use npx skills add bradautomates/claude-video -g for global installation
• claude.ai web users must download the watch.skill bundle from GitHub releases and upload via Settings
• Developers can clone the repo and symlink skills/watch to the host's skill directory
• All platforms share the identical code in skills/watch/ with SKILL.md as the universal contract

Frequently Asked Questions

Does Claude Video work on Windows?

Yes. Because the skill uses pure Python standard library scripts and external tools (ffmpeg, yt-dlp) that support Windows, it functions on Windows through Claude Code, Cursor, or any Agent Skills host with Windows support. The npx skills CLI and manual symlink methods work on Windows when using appropriate path formats.

Can I install Claude Video without Node.js?

Yes. While the npx skills method requires Node.js for Codex/Cursor installations, you can avoid Node.js entirely by using the Claude Code plugin marketplace commands or the manual installation method. Simply clone the repository and symlink the skills/watch folder to your host's skill directory.

What dependencies does Claude Video require?

The skill requires ffmpeg for video processing and yt-dlp for downloading videos from URLs. These are checked by skills/watch/scripts/setup.py on first run. Optional Whisper API support requires configuring an API key in your environment. The Python scripts themselves use only standard library modules.

How do I update Claude Video?

Update methods vary by platform:

• Claude Code: Run /plugin update watch@claude-video
• Codex/Cursor/etc.: Re-run npx skills add bradautomates/claude-video -g to pull the latest version
• Web/manual: Download the latest watch.skill bundle or pull the Git repository and restart your host

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}