Using ffmpeg command line to achieve an mcp server, can be very convenient, through the dialogue to achieve the local video search, tailoring, stitching, playback and other functions The server implements the following tools: - findvideopath The parameters are directory and file name, file name can be complete, or is not suffixed, recursive search in the directory, return the full path - getvideoi
Add this skill
npx mdskills install video-creator/ffmpeg-mcpComprehensive video processing toolkit with 8 FFmpeg-based tools, but platform-limited
Using ffmpeg command line to achieve an mcp server, can be very convenient, through the dialogue to achieve the local video search, tailoring, stitching, playback and other functions
The server implements the following tools:
find_video_path
The parameters are directory and file name, file name can be complete, or is not suffixed, recursive search in the directory, return the full path
get_video_info
The parameters are video path, return the video info, linkes duration/fps/codec/width/height.
clip_video
The parameter is the file path, start time, end time or duration, and returns the trimmed file path
concat_videos
The parameters are the list of files, the output path, and if the video elements in the list of files, such as width, height, frame rate, etc., are consistent, quick mode synthesis is automatically used
play_video
Play video/audio with ffplay, support many format, like mov/mp4/avi/mkv/3gp, video_path: video path speed: play rate loop: play count
overlay_video
Two video overlay.
background_video: backgroud video path
overlay_video: front video path
output_path: output video path
position: relative location
dx: x offset
dy: y offset
scale_video
Video scale.
video_path: in video path
width: out video width, -2 keep aspect
height: out video height, -2 keep aspect
output_path: output video path
extract_frames_from_video
Extract images from a video.
Parameters:
video_path (str): The path to the video.
fps (int): Extract one frame every specified number of seconds. If set to 0, extract all frames; if set to 1, extract one frame per second.
output_folder (str): The directory where the images will be saved.
format (int): The format of the extracted images; 0: PNG, 1: JPG, 2: WEBP.
total_frames (int): The maximum number of frames to extract. If set to 0, there is no limit
More features are coming
git clone https://github.com/video-creator/ffmpeg-mcp.git
cd ffmpeg-mcp
uv sync
{
"mcpServers": {
"ffmpeg-mcp": {
"autoApprove": [],
"disabled": false,
"timeout": 60,
"command": "uv",
"args": [
"--directory",
"/Users/xxx/Downloads/ffmpeg-mcp",
"run",
"ffmpeg-mcp"
],
"transportType": "stdio"
}
}
}
Note: the value:/Users/XXX/Downloads/ffmpeg in args need to replace the actual download ffmpeg-mcp directory
Currently, only macos platforms are supported, including ARM64 or x86_64
Install via CLI
npx mdskills install video-creator/ffmpeg-mcpFfmpeg MCP is a free, open-source AI agent skill. Using ffmpeg command line to achieve an mcp server, can be very convenient, through the dialogue to achieve the local video search, tailoring, stitching, playback and other functions The server implements the following tools: - findvideopath The parameters are directory and file name, file name can be complete, or is not suffixed, recursive search in the directory, return the full path - getvideoi
Install Ffmpeg MCP with a single command:
npx mdskills install video-creator/ffmpeg-mcpThis downloads the skill files into your project and your AI agent picks them up automatically.
Ffmpeg MCP works with Claude Code, Claude Desktop, Cursor, Vscode Copilot, Windsurf, Continue Dev, Gemini Cli, Amp, Roo Code, Goose. Skills use the open SKILL.md format which is compatible with any AI coding agent that reads markdown instructions.