# KyaniteLabs/kinocut [Health: Active]

**Category:** 🎥 Multimedia Process  
**Repository:** https://github.com/KyaniteLabs/kinocut  
**GitHub Stars:** 142  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/kyanitelabs-kinocut

## Description
Guardrailed video editing for AI agents with 135 MCP tools for FFmpeg editing, captions, audio, effects, Hyperframes, repurposing, resumable workflows, quality gates, and provenance receipts. Includes a Python client and kino CLI; runs locally with no required API keys.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "kinocut": {
    "command": "npx",
    "args": ["-y","--yes"]
  }
}
```

## Documentation

## What KyaniteLabs/kinocut MCP server does

KyaniteLabs/kinocut MCP server gives AI agents a structured interface for editing local video instead of requiring them to construct raw FFmpeg commands. The project includes an MCP server, a Python client, and the `kino` command-line interface. Its editing surface covers common operations such as trimming, resizing, subtitle rendering, audio normalization, effects, and export workflows.

The main use case is turning longer local recordings into reviewable short-form media. An agent can select a segment, generate or apply captions, change the aspect ratio for vertical output, and run a release checkpoint before the result is published. The project also describes workflows for podcasts, interviews, Shorts, Reels, TikTok packages, and 360 dual-camera assembly.

## How it works

The KyaniteLabs/kinocut MCP server runs on the user’s machine and invokes FFmpeg for media processing. Typed operations, preflight validation, and quality or release checkpoints provide structure around those transformations. Outputs can be accompanied by Video Receipts containing provenance and review information, allowing an agent or human operator to inspect how an asset was produced.

A Python client is available for scripted workflows. For example, a workflow can trim a source file, transcribe it, render a subtitle file, resize the result to a 9:16 frame, and call a release checkpoint. Resumable workflows and manifests are also part of the documented editing model, which suits repeatable jobs in scripts or CI environments.

Optional capabilities include Hyperframes and Whisper-related extras. The repository also documents an approval-based 360 assembly path that creates a reviewable plan before rendering the chosen layout with FFmpeg.

## Setup and configuration

Install the published Python package with `pip install kinocut`, then start the MCP mode with `kino --mcp`. Python 3.11 or newer is listed by the project. FFmpeg is a runtime requirement and must be discoverable on the system `PATH`.

The server is local-first and supports macOS, Linux, and Windows according to the README. It does not require an API key. The repository distinguishes the stable PyPI release from development code in the source repository, so production workflows should use the published package when a released version is preferred.

## Tools and capabilities

The documented tool surface includes capabilities for:

- Trimming and transforming local video files.
- Generating or applying captions and subtitles.
- Normalizing or otherwise processing audio.
- Adding video effects and changing output dimensions or aspect ratios.
- Repurposing recordings into short-form vertical media.
- Running quality gates and release checkpoints.
- Producing provenance-oriented Video Receipts and review artifacts.
- Supporting resumable workflows and 360 dual-camera assembly.

The repository describes a larger MCP and CLI surface that changes between published releases and development tip, so exact tool counts depend on the version installed.

## Limitations and notes

KyaniteLabs/kinocut MCP server is not a hosted cloud editor or a credit-metered SaaS service. Processing is performed against local media on the machine running the server. FFmpeg is required separately, and some features depend on optional Hyperframes or Whisper extras.

The project is designed for guarded, inspectable workflows; it does not replace human approval when a release checkpoint is part of the process. The README names Claude Code and Cursor among its intended MCP clients and states that it works with MCP clients generally, but client-specific setup details may vary. The source is released under Apache-2.0.

_Full upstream README: https://allmcps.com/mcp/kyanitelabs-kinocut/readme_