CartCut MCP Server
io.github.cartesiancs/cartcut
AI-powered video editor with transcript-based cutting, captions, effects, and Claude integration via MCP.
What is the CartCut MCP server?
CartCut is an AI video editor designed for motion effects and versatility. It exposes its live timeline over MCP, enabling Claude Code and other AI agents to edit projects by cutting from transcripts, adding captions, trimming, applying motion, transitions, and effects. Edits integrate into the native undo history.
CartCut is a desktop video editor optimized for AI-assisted editing through MCP. It combines traditional video editing features—cut editing, audio mixing, layer-based composition, keyframe animation—with AI capabilities like auto-captioning via Whisper and Claude Code integration. Use it to rapidly produce videos with AI-generated edits, motion graphics, and effects without heavy software like After Effects.
How to install CartCut
Copy-paste configuration for popular MCP clients.
CARTCUT_MCP_TOKENsecretOptional. CartCut's MCP token; read from CartCut's own settings when unset.
Tools & capabilities
Tools this server exposes to the agent.
Cut Edit— Trim and cut video segments from transcriptsAudio Mixing— Mix and balance audio tracksKeyframe Animation— Create animations with keyframe-based motion, scale, opacity, and rotationText Editing— Add and edit text overlaysAI Auto Caption— Generate captions automatically using WhisperEffects and Transitions— Apply visual effects, transitions, blur, and chromakeyLayer-based Editing— Unlimited layers for flexible asset composition and effect applicationProject Management— Save and load projects as filesScreen & Audio Recording— Record screen and audio directly in the editorShape Drawing— Draw shapes and vector graphicsMulti-resolution Export— Support for 8K, 4K, and other resolutions with FFmpeg rendering
Use cases
- Generate video summaries by having Claude cut key moments from a transcript and add captions
- Automate motion graphics creation with AI-assisted keyframe animation and effects
- Produce social media content with rapid AI-driven editing, trimming, and caption generation
- Create educational videos with auto-generated captions and AI-assisted scene arrangement
- Batch edit multiple videos by scripting cuts, transitions, and effects through Claude Code
CartCut MCP server FAQ
CartCut is a desktop video editor that combines traditional editing tools (cutting, audio mixing, effects) with AI capabilities. It integrates with Claude Code and other AI agents via MCP to enable AI-assisted editing of your live timeline.
CartCut is open-source under the MIT license. The desktop application is available for download on macOS (with Windows support in development).
Install the plugin via `/plugin marketplace add cartesiancs/cartcut` and `/plugin install cartcut-editing@cartcut`, then open CartCut, click the ⚡ icon, and copy the connection command with your MCP token.
Use the MCP bridge package: `npx -y @cartesiancs/cartcut-mcp`. It automatically finds your CartCut token. Configure it in your MCP settings for Cursor or Claude Desktop.
CartCut supports all standard video and audio formats including mp4, mov, mp3, and wav. It uses FFmpeg for fast rendering and supports resolutions up to 8K.
CartCut requires an MCP token (a UUID) generated in the desktop app's connection panel (⚡ icon). This token authenticates the AI agent to edit your live project.
README (reference)
Source of truth, from the repository.


<a href="https://cartesiancs.com/cartcut"><img alt="Download for macOS" src="https://img.shields.io/badge/Download_for-macOS-000000?style=for-the-badge&logo=apple&logoColor=white" /></a> <a href="#"><img alt="GitHub Repo stars" src="https://img.shields.io/github/stars/cartesiancs/cartcut?style=for-the-badge" /></a> <a href="#"><img alt="GitHub Repo stars" src="https://img.shields.io/github/license/cartesiancs/cartcut?style=for-the-badge" /></a> <img alt="GitHub Release" src="https://img.shields.io/github/v/release/cartesiancs/cartcut?style=for-the-badge">
</p> <p align='center'> English | <a href="./docs/README.ko.md">한국어</a> </p> <p align='center'> <a href="https://www.youtube.com/watch?v=Bh06VOYSMIM">View Demo</a> · <a href="https://github.com/cartesiancs/cartcut/issues">Report Bugs</a> · <a href="https://github.com/cartesiancs/cartcut/releases"><b>Download</b></a> · <a href="https://github.com/cartesiancs/cartcut/issues/new">Suggest Features </a> </p>Video editing software designed for motion effects and versatility.
In addition to essential features like basic cut editing, animation, sound mixing, external library extensions, project management, and text editing, our software offers a wide range of powerful tools.
It also supports layer-based editing, which differs from traditional track-based editing. This approach makes it easier to apply multiple effects to individual assets, providing greater flexibility and creative control.
About The Project
You can check out a limited demo of the website at the following link.
Features
- Cut Edit
- Support for all standard formats (mp4, mov, mp3, wav...)
- Audio mixing
- Fast rendering with FFmpeg
- Unlimited layers
- Cross Platform
- Re-position, Scale, Opacity, Rotation animation, Keyframe
- Add Text
- External Extension
- Save&Load Project as File
- Multilingual Support
- 8k Edit & 4k Edit & more resolution
- Screen Record & Audio Record
- Chromakey
- AI Auto Caption (whisper)
- Blur Effect (WebGL)
- Draw shape
- Effects and Transitions
- and more...
Editing with Claude Code
Cartcut exposes its live timeline over MCP, so Claude Code can edit the project you are looking at: cutting from a transcript, adding subtitles, trimming, restaging clips, motion, transitions and effects. Its edits land in your own undo history, so ⌘Z takes them back.
Install the plugin, which carries both the editing skill and the connection:
/plugin marketplace add cartesiancs/cartcut
/plugin install cartcut-editing@cartcut
It asks for your Cartcut MCP token once. Open Cartcut, click the ⚡ icon at the
bottom right, and copy the connection command; the token is the UUID after
Bearer . Then ask for an edit in plain language.
The skill is also on the skills CLI, for Claude Code and for other agents:
npx skills add cartesiancs/cartcut --skill cartcut-editing -a claude-code -g
That route installs the skill alone, so run the ⚡ panel's claude mcp add line
afterwards to connect the bridge. Take one route or the other, not both.
Any other MCP client (Claude Desktop, Cursor, Codex) connects through
@cartesiancs/cartcut-mcp, which is also listed in the official MCP
Registry. It finds the token by itself:
npx -y @cartesiancs/cartcut-mcp
plugins/cartcut-editing/README.md has the details, including what to do when it will not connect.
Installation
First, install dependencies.
npm install
and, Download ffmpeg and ffprobe into ./bin, in a folder named for the
target you are building. Only the folder matching your machine is needed to run
the app locally; npm run build reads whichever one it is packaging for.
bin/
darwin-arm64/{ffmpeg,ffprobe} Apple Silicon
darwin-x64/{ffmpeg,ffprobe} Intel Mac
win32-x64/{ffmpeg.exe,ffprobe.exe}
Compatible binaries can be downloaded from https://github.com/cartesiancs/ffmpeg4nugget
The macOS builds must be native. An x86_64 binary runs on Apple Silicon
through Rosetta 2 at roughly half the export speed, and nothing in the app will
say so. Check with lipo -archs bin/darwin-arm64/ffmpeg, which must print
arm64. The build also has to carry libx264, libx265, libvpx-vp9,
prores_ks and the VideoToolbox encoders; ffmpeg -encoders lists them.
next, Permission grant is required. Please enter the command below to grant permission for bin folder.
chmod -R 777 bin
Running
npm run dev
npm run start
Releasing
npm run build:osx signs, notarizes and uploads the macOS build to a draft GitHub release. Publishing that draft runs mirror-r2.yml, which copies the release to Cloudflare R2 and updates the manifest the website's download button reads:
https://download.cartesiancs.com/cartcut/latest.json: version plus the Apple Silicon and Intel.dmgURLshttps://download.cartesiancs.com/cartcut/latest.txt: the version alone, as plain text
The GitHub release is the source of truth, and in-app auto-update still reads it. To mirror a tag that was published before the workflow existed: gh workflow run mirror-r2.yml -f tag=v0.5.3.
Roadmap
Our ultimate goal is to empower creators to produce motion graphics effortlessly. We hope they can achieve stunning motion effects without relying on heavy software like After Effects.
Please refer to the roadmap file for more details.
Contributors
<a href = "https://github.com/cartesiancs/cartcut/graphs/contributors"> <img src = "https://contrib.rocks/image?repo=cartesiancs/cartcut"/> </a>License
We are adopting the MIT license. license file
Related MCP servers

Inlay
Make any website discoverable by AI agents via MCP, llms.txt, and structured data.
Vehicle specs, federal recalls, complaints, TSBs, investigations and OBD-II DTC data for AI agents.

JSON-first Coreutils CLI for AI agents. 114 commands with sandbox, dry-run, and MCP support.

io.github.cashfree/cashfree-mcp
AI-powered integration with Cashfree payment gateway, payouts, and identity verification APIs.
Solana memecoin rug + creator-reputation checks for agents. PASS/CAUTION/FAIL + 0-100 score.

Privacy-first file tools for AI agents — PDF & image ops, metadata scrubbing, background removal.
