What it does
The server exposes FFmpeg and FFprobe to MCP-compatible AI tools through 19 typed tools defined with Zod schemas. The AI passes parameters such as an input path and a preset, and the server builds and runs the FFmpeg command locally, with error handling. The tools cover probing media metadata, converting formats, trimming, compressing to a target file size, extracting audio and thumbnails, and generating waveforms and contact sheets.
It is suited to tasks such as reframing landscape video to 9:16 with blurred backgrounds, burning in or embedding subtitles, stacking split-screen reels, normalising loudness to EBU R128 (-16 LUFS), removing silent pauses while keeping lip-sync, and extracting a 16kHz mono WAV for Whisper speech recognition. Image previews are returned as base64 so vision models can inspect thumbnails, waveforms and keyframe grids.
Running it requires FFmpeg and FFprobe to be available to the server, as all processing happens locally. Hardware encoding is detected automatically for VideoToolbox on macOS and for NVIDIA NVENC, Intel QuickSync, AMD AMF and VA-API on Linux and Windows. Progress is reported through standard MCP progress notifications, and an AbortSignal from the client stops the child FFmpeg processes.
Limitations and requirements
Processing runs locally, so FFmpeg and FFprobe must be available to the server. Hardware acceleration applies only to the encoders listed for each platform (VideoToolbox, NVENC, QuickSync, AMF, VA-API).
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 18 Nov 2026 · its page on the source
FFmpeg MCP Server
FFmpeg & FFprobe media processing toolkit for video, audio, and social formats
v0.6.2As shown on registry.modelcontextprotocol.io when we last read it