Skip to content
sourceon GitHub →
for Claude Code · Cursor · Claude Desktop

Convert files via MCP

A local MCP server that plugs the file converter directly into your AI assistant. Say "convert X to Y" in chat — it runs ffmpeg locally, returns the path. Your files never leave your machine.

Drop this in your MCP config
{
  "mcpServers": {
    "converter": {
      "command": "npx",
      "args": ["-y", "@sergeyfaraday/converter-mcp@latest"]
    }
  }
}
Config locations · Cursor: ~/.cursor/mcp.json · Claude Code: .mcp.json · Claude Desktop: claude_desktop_config.json
How it works
  1. 1
    Install
    Add one block to your MCP config file. Your AI client will auto-install on first run via npx.
  2. 2
    Ask
    Type something like "convert ~/video.mov to webm" in your chat. Your assistant picks the right tool automatically.
  3. 3
    Get the file
    The converted file appears next to the original in seconds. Your AI returns the path — click to open.
Tools exposed
convert_file
Convert a local file. Takes input, target, optional output and quality.
list_formats
Returns every supported format and conversion pair so the model picks the right target.
Supported formats (230 pairs)
Image
.png · .jpg · .webp · .gif · .bmp · .tiff · .avif* · .heic* · .svg* · .ico* · .pdf*
Video
.mp4 · .webm · .mov · .avi · .mkv · .flv · .3gp · .ts
Audio
.mp3 · .wav · .ogg · .m4a · .flac · .aiff · .aac
* decode-only (input)
FAQ

Model Context Protocol is an open standard that lets AI assistants talk to external tools. This package is a local MCP server — your AI client (Claude Code, Cursor, Claude Desktop, etc.) launches it on your machine and can call its conversion tool in any chat.

No. The MCP server runs entirely on your computer, spawns the bundled native ffmpeg binary, and writes the output to your disk. The server has no network calls.

Anything that supports MCP over stdio — Claude Code, Cursor (via `.cursor/mcp.json`), Claude Desktop (via `claude_desktop_config.json`), Zed, Continue, Cline, and any other client that reads the standard `mcpServers` config block.

About 5–10× faster on video and audio because it uses the native ffmpeg binary instead of WebAssembly. A 1080p 30-second WebM → MP4 finishes in ~4 seconds on a laptop.

All the same formats as the web app: PNG, JPG, WebP, AVIF, GIF, BMP, TIFF, HEIC, SVG, PDF (→ raster), MP4, WebM, MOV, AVI, MKV, FLV, 3GP, TS, MP3, WAV, OGG, M4A, FLAC, AIFF, AAC — 230 conversion pairs in total.

Drop the snippet on this page into your client's MCP config, restart the client, and ask the AI to convert a file. The first call pulls the package via npx; subsequent calls are cached.

Prefer drag & drop?
Use the browser version →