ffmpeg -lite
MCP server for video/audio processing via FFmpeg - convert, compress, trim, extract audio, add subtitles
claude mcp add --transport stdio kevinwatt-ffmpeg-mcp-lite uvx ffmpeg-mcp-lite \ --env FFMPEG_PATH="Path to ffmpeg binary (default: ffmpeg)" \ --env FFPROBE_PATH="Path to ffprobe binary (default: ffprobe)" \ --env FFMPEG_OUTPUT_DIR="Default output directory (default: ~/Downloads)"
How to use
This MCP server provides FFmpeg-based media processing capabilities exposed as MCP tools with the ffmpeg_ prefix. It can retrieve media information, convert and compress formats, trim videos, merge clips, extract audio and frames, and burn subtitles. Tools include ffmpeg_get_info, ffmpeg_convert, ffmpeg_compress, ffmpeg_trim, ffmpeg_merge, ffmpeg_extract_audio, ffmpeg_extract_frames, and ffmpeg_add_subtitles. Each tool accepts parameters described in the README (for example file_path, output_format, quality, start_time, end_time, and more). To use them from your MCP client, reference the tools by their names and provide the required arguments in natural language or structured requests. The server integrates with Claude, Dive, Windsurf, VS Code Copilot, and other MCP clients, enabling natural language prompts or CLI-style commands to orchestrate complex media workflows end-to-end. Begin by ensuring FFmpeg is installed and visible to the MCP server, then call the appropriate ffmpeg_* tool with the needed parameters to perform the task you described (e.g., metadata extraction, format conversion, or frame extraction).
How to install
Prerequisites:
- Python 3.10 or newer (if using the Python package)
- FFmpeg installed and accessible in your system PATH
- Access to the internet to download the MCP package
Installation (Python package):
pip install ffmpeg-mcp-lite
Installation (uv/virtual environment alternative):
uv pip install ffmpeg-mcp-lite
Configuration (example):
- Ensure FFmpeg is installed and FFmpeg/FFprobe binaries are discoverable (FFMPEG_PATH and FFPROBE_PATH can be set via environment variables).
- Add the MCP server config to your client (see mcp_config section in this document).
Verification:
- After installation, verify that the server is reachable from your MCP client by sending a basic metadata request using ffmpeg_get_info on a sample file.
- Check logs if the server fails to initialize, and confirm that the uvx command is resolving to the ffmpeg-mcp-lite package.
Additional notes
Environment variables can customize where FFmpeg binaries are located and where outputs are written: FFMPEG_PATH, FFPROBE_PATH, and FFMPEG_OUTPUT_DIR. If using uvx or a virtual environment, ensure the environment is active when starting the MCP client and that the ffmpeg-mcp-lite package is installed in that environment. Common issues include FFmpeg not found in PATH, incorrect file_path formats, or missing permissions for output directories. The tools support a range of FFmpeg operations; consult the available tool descriptions to supply required parameters (e.g., duration formats like 00:01:30 or seconds). When integrating with Claude or Dive, you can embed these tool calls in natural language prompts to orchestrate multi-step media workflows (metadata, conversion, trimming, and subtitle burning) in a single session.
Related MCP Servers
MCP-Bridge
A middleware to provide an openAI compatible endpoint that can call MCP tools
asterisk
Asterisk Model Context Protocol (MCP) server.
skill-to
Convert AI Skills (Claude Skills format) to MCP server resources - Part of BioContextAI
google-search-console
It connects directly to your Google Search Console account via the official API, letting you access key data right from AI tools like Claude Desktop or OpenAI Agents SDK and others .
fcpxml
🎬 The first AI-powered MCP server for Final Cut Pro XML. Control your edits with natural language.
whatsapp -extended
Extended WhatsApp MCP server with 41 tools - reactions, group management, polls, presence, newsletters & more. Fork of lharries/whatsapp-mcp