ffmpeg-skill: Local Video Editor and MCP Server for AI Coding Agents
GitHub Open Source
Equip Claude Code or Cursor with 42 local FFmpeg tools that cut, caption, and normalise video without cloud APIs.
Repository snapshot
- Repository: kajisho5/ffmpeg-skill
- Language: Python
- License: MIT
- Stars: 4 424
- Forks: 349
Daily open-source picks: GitHub Open Source on Telegram
What it is
ffmpeg-skill is an agent skill and MCP server designed for Claude Code, Cursor, and Codex. It bridges coding agents directly to local FFmpeg and Python standard libraries, giving them 42 structured tools to handle video processing tasks locally without uploading files or using cloud API keys.
Why it matters
When AI agents try to edit video using raw shell strings, they often guess codecs, assume frame rates, and fail silently. This project replaces guesswork with a structured workflow: probe the input file first, edit using typed arguments, and verify the output using local checks and visual contact sheets.
Key features
- Run 42 local video tools covering cuts, silence removal, captions, karaoke, and loudness normalisation
- Expose an MCP server with a machine-readable contract for Claude Code, Cursor, and Codex
- Verify every execution automatically using built-in probing, checks, and visual contact sheets
- Operate fully offline without cloud dependencies, subscriptions, or external API keys
- Generate runtime input schemas directly from the CLI argparse definitions to prevent contract drift
Quick start
npx ffmpeg-skillWho should try it
This tool fits developers building local media pipelines, automated video generation workflows, or leveraging AI coding assistants for local file manipulation.
Verdict
An essential utility for anyone building local AI media workflows, provided you already have FFmpeg and Python installed.
Continue exploring
Get the next pick in Telegram: Follow GitHub Open Source
Curated from the repository README and metadata. No hands-on test is claimed; evaluate it in a sandbox before production use.