Log inSubmit server
MCP Video logo

MCP Video

Local MCP video editing with FFmpeg, structured tools, quality checks, review checkpoints, and safe output handling.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install MCP Video

Pick your client, then add it in one click or paste the config.

STDIOPython
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "mcp-video": {
      "command": "uvx",
      "args": [
        "--from",
        "kinocut==2.0.1",
        "kino"
      ]
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Automated check passed— started and listed 18 tools correctly ().

Overview

mcp-video, now distributed as Kinocut, gives AI agents local tools for editing and inspecting video through FFmpeg. It exposes a compact default tool set, with additional editing, effects, audio, AI, image, quality-control, planning, and platform operations available through profiles or dynamic discovery. Edits are validated before execution, normally write to new files, and return structured results that agents can inspect. Reach for it when an agent needs to trim, caption, resize, package, review, or quality-check real media without uploading files or publishing exports.

Use cases

  • Trim interviews into captioned vertical clips
  • Prepare podcast footage for Shorts, Reels, and TikTok
  • Run quality gates before releasing a video
  • Build resumable multi-step editing workflows

Key features

  • FFmpeg-backed local video editing
  • Dynamic MCP tool discovery with profiles
  • Preflight validation and non-overwriting outputs
  • Structured results and Video Receipts
  • Quality checks for frames, silence, and loudness
  • Review cards, storyboards, and release checkpoints

Tools18

What MCP Video lets your agent do. Expand a tool for its parameters.

Verified live~4.3k tokens of context

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked

Recent health check succeeded.

25/30 checks healthy over the last 95d

What it can do

From the 18 tools this server returned when we ran it; 18 declare their own safety hints.

Can delete or modify data · 14Makes network requests · 5Read-only tools · 4
Network at startup2 connections
Files written at startup10

Observed 2026-11-13 in an isolated sandbox. A description of reach, not a security audit.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 6d ago via OSV.dev · kino (PyPI)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars264Stargazers on the source repository.
Last commit1d agoMost recent push to the default branch.
Availability100%Our rolling endpoint + install checks that succeeded.
Install checkPassedOur sandbox started it and listed its tools.
Tools exposed18Callable tools this server registers over MCP.
Directory activity6 viewsConfig copies, upvotes, and views on AllMCPs.

How MCP Video works

What mcp-video does

Kinocut is the renamed distribution of mcp-video: a local video-editing MCP server with a Python client and the kino command-line interface. It uses FFmpeg to process real media on the machine where it runs. The package supports common editing tasks such as trimming, merging, captioning, resizing, converting, overlays, cropping, speed changes, audio work, and frame extraction.

Beyond individual edits, it can create vertical shorts from selected ranges, prepare podcast or stream clips for Shorts, Reels, and TikTok, run release checkpoints, and execute multi-step jobs from JSON specifications. Hyperframes and Revideo tools cover code-driven motion graphics. Other operation groups address quality control, video planning, AI-assisted processing, images, platform packaging, and audio mixing.

How it works

The MCP interface offers a small default catalog to reduce the amount of tool metadata an agent must read. That default exposes 18 entries, including eight frequently used tools, eight operation families, list_operations, and describe_operation. Profiles can expose tools for a particular job, while KINOCUT_TOOLS=all makes the full published catalog available.

Operation families provide access to the broader catalog without listing every tool individually. An agent can enumerate a family with list_operations, retrieve a specific schema with describe_operation, and then dispatch the operation. Family dispatch uses the same implementation as the corresponding named tool.

Before FFmpeg runs, inputs such as filter ranges, merge compatibility, audio levels, overlay opacity, and text timing are checked. Outputs use new paths by default; replacing an existing file requires an explicit overwrite option. Results are structured for agent inspection, and Video Receipts record the request, executed work, and changes. Quality gates can detect issues including black frames, silence, and unsuitable loudness.

Setup and configuration

FFmpeg must be installed and available on PATH. The server can be started through an MCP client's JSON configuration with uvx, for example by running kino --mcp from the kinocut package. It can also be installed with pip, after which kino doctor reports available features and requirements. Optional capabilities are installed as package extras.

Claude Code can register the server with its MCP command. JSON-based configuration is supported for Claude Desktop, Cursor, Windsurf, and Zed. A local Claude Desktop extension can be built as an unsigned .mcpb, but that package still launches the user's Python environment, requires Node and FFmpeg, and is not a sandbox.

The KINOCUT_TOOLS setting controls catalog exposure. It can select a named profile or expose all tools; leaving it unset uses the compact default listing.

Tools and capabilities

The hot-path tools include:

  • video_info for media inspection
  • video_trim and video_merge for basic edits
  • video_add_text and video_add_audio for overlays and sound
  • video_extract_frame for still extraction
  • video_shorts for short-form cuts
  • video_doctor for diagnostics

Additional families cover video editing, effects, audio, AI video operations, image processing, quality control, planning, and platform tasks. video_review_card creates an HTML review card with a contact sheet, media details, and approval or revision actions. kino release-checkpoint produces a thumbnail, storyboard, and pass/fail quality report beside a clip.

Limitations and notes

Processing is local, and the server does not post or publish exports. Optional model features remain disabled until enabled, with controls for vision and network behavior. The package renders and packages existing media rather than generating video from text or providing a paid image-generation adapter. Caption speech uses local eSpeak NG rather than a neural voice. The kinocut_sound program includes a limited S12 join, and its benchmark uses synthetic clips rather than full-episode or human-listening approval.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about MCP Video

mcp-video is an MCP server that connects AI agents to local FFmpeg-based video workflows through Kinocut. Its main tools inspect media, trim and merge clips, add text or audio, extract frames, create short-form edits, and run video diagnostics, with additional operations available through tool families and profiles.

Install FFmpeg on your PATH, then register Kinocut with an MCP client using `uvx --from kinocut kino --mcp`, or install the package with `pip` and run `kino doctor`. Optional features may require package extras.

Yes, mcp-video can be configured in Cursor through its JSON MCP server configuration. The same setup pattern also supports Claude Desktop, Windsurf, and Zed.

mcp-video writes edits to new files by default and requires an explicit overwrite option to replace an existing file. It also validates edit parameters before invoking FFmpeg and returns structured results for inspection.

No, mcp-video renders and packages media but does not publish it. Its platform-related operations can prepare packages for destinations such as Shorts, Reels, and TikTok, leaving publication to the user.

Related MCP servers

All in Multimedia Process All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewMCP Video AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/mcp-video?style=directory)](https://allmcps.com/mcp/mcp-video)
HTML Embed
<a href="https://allmcps.com/mcp/mcp-video"><img src="https://allmcps.com/api/badge/mcp-video?style=directory" alt="MCP Video on AllMCPs" /></a>