Automate Video Compression with the CLI, MCP Server & AI Agents

July 2, 2026 • VideoRecompress Studio Team

Most people meet VideoRecompress Studio through its Windows wizard: drop in a folder of videos, pick a preset, click Start. But the wizard is only one way to drive it. The same compression engine — H.264, H.265, AV1, VP9, GPU hardware encoding — also ships with a full command-line interface called videorecompress, a built-in MCP server, and an installable Claude Code skill.

This post walks through all three: scripting with the videorecompress CLI, wiring the MCP server into Claude Desktop or Claude Code, and letting an AI agent run your compression pipeline from plain English. Full CLI and MCP reference lives in the public GitHub repository at github.com/contentasoftware/videorecompress-studio-plugin

Three ways to automate

• Command line

  Scripts, scheduled tasks and batch runs.

• MCP server

  AI agents call VideoRecompress as a tool.

• Claude Code skill

  Claude Code learns the commands from your skills folder.

The videorecompress CLI: the same engine, scriptable

The executable is called videorecompress and installs together with the desktop app — same installer, same license. A default install puts it in %LOCALAPPDATA%\Programs\VideoRecompressStudio\ and adds that folder to your PATH, so videorecompress works from any new terminal.

It exposes 9 commands: analyze, recompress, batch, presets, profile, watch, status, register, and serve. Analyze a file, recompress with codec and CRF settings, or batch-process an entire folder — all from one line:

videorecompress info video.mp4 --json

videorecompress recompress video.mp4 --output ./compressed --preset phone_archive

videorecompress batch ./raw-footage --output ./archived --preset wedding_archive --workers 2

Every command accepts --json for machine-readable output (JSON or NDJSON for batch progress) and returns proper exit codes, so the CLI drops cleanly into PowerShell scripts, bash pipelines, CI jobs, and scheduled tasks.

Two preset flags. --preset phone_archive selects a named preset (codec, CRF and expected savings). --encoder-preset medium sets the encoder speed (ultrafast through veryslow). Use --preset for the named presets and --encoder-preset only when tuning encoder speed.

Real automation recipes

1. The scheduled archive job

Point batch at a folder of raw footage and let it recompress overnight with a quality-first preset. One command turns a memory-card dump into an archived library:

videorecompress batch ./raw-footage --output ./archived --preset wedding_archive --workers 2

Schedule it in Task Scheduler to run weekly and your camera imports compress automatically — original files untouched, H.265 archives ready for long-term storage.

2. The self-running watch folder

The watch command monitors a folder and recompresses every new video automatically. Start it once and it keeps running — drop files in, get compressed files out:

videorecompress watch C:\Videos --preset phone_archive --output C:\Compressed

Run it at logon via Task Scheduler and you have a zero-click compression service: phone dumps, screen recordings, and camera imports land in the incoming folder, and smaller H.265 files appear in the output folder seconds later.

The MCP server: hand the engine to an AI agent

MCP (Model Context Protocol) is the open standard AI assistants use to call external tools. VideoRecompress ships with a built-in MCP server — you start it with a single command:

// claude_desktop_config.json
{
  "mcpServers": {
    "videorecompress": {
      "command": "videorecompress",
      "args": ["serve"]
    }
  }
}

It exposes 5 tools — analyze video, recompress a file, batch recompress, list presets, and estimate space savings — and it works during the free trial. To connect Claude Desktop, add this to your claude_desktop_config.json:

From then on you can ask your assistant things like “recompress every MP4 in my Downloads folder with the phone_archive preset” and it calls the local engine directly. Your videos never leave your machine — the agent orchestrates, the compression runs locally.

Claude Code skill: video compression in plain English

If you use Claude Code, the /contenta-video skill ships inside the VideoRecompress Studio plugin together with the MCP server, so one install gives the agent both. Two commands:

/plugin marketplace add contentasoftware/videorecompress-studio-plugin
/plugin install videorecompress-studio@videorecompress-studio-plugin

In Cursor, Claude Desktop, VS Code or any other MCP client, add this server entry (needs Node.js 18 or newer):

{
  "mcpServers": {
    "videorecompress-studio": { "command": "cmd", "args": ["/c", "npx", "-y", "@contentasoft/videorecompress-mcp"] }
  }
}

Then just describe the job in your own words — “batch recompress this wedding folder with max quality” or “watch my screen recordings folder and compress new files”. The skill translates the request into the right videorecompress commands and reports back.

Everything is documented on GitHub

The plugin source, the MCP server reference and the CLI documentation live in the public GitHub repository, and the launcher package is published on npm: github.com/contentasoftware/videorecompress-studio-plugin · npmjs.com/package/@contentasoft/videorecompress-mcp

The docs are maintained alongside the product, so what you read there matches the binary on your disk. Copy the recipes straight into your own scripts, or open an issue if you hit an edge case.

Frequently Asked Questions