video-editor-mcp
An MCP server that provides video editing tools via ffmpeg, enabling tasks like trimming, merging, adding overlays, and format conversion.
README
video-editor-mcp
An MCP (Model Context Protocol) server that gives Claude (or any MCP-compatible
client) real video-editing tools, powered by ffmpeg. Once connected, you can
ask Claude things like "trim intro.mp4 to the first 10 seconds, add my logo
in the corner, and export it as a GIF" and it will call these tools directly.
Requirements
- Node.js 18+
- ffmpeg and ffprobe installed and available on your system
PATH- macOS:
brew install ffmpeg - Ubuntu/Debian:
sudo apt install ffmpeg - Windows: ffmpeg.org/download.html, then add it to PATH
- macOS:
Install
cd video-editor-mcp
npm install
Configure where your videos live
Every tool takes file paths. Relative paths are resolved against the
MEDIA_DIR environment variable (so Claude doesn't need to know your full
filesystem layout) — set it to the folder containing the videos you want to
edit. Absolute paths are always used as-is. If MEDIA_DIR isn't set, it
defaults to the directory you launch the server from.
Connect it to an MCP client
This is a standard MCP server using the stdio transport, so any MCP-compatible client/host can spawn and talk to it. The host just needs to launch this process and communicate over stdin/stdout:
- Command:
node - Args:
["/absolute/path/to/video-editor-mcp/src/index.js"] - Env:
MEDIA_DIR=/absolute/path/to/your/videos
How you wire that up depends on your host/client's own configuration format
(most MCP-aware hosts use a command + args + env shape similar to the
above).
Tools exposed
| Tool | What it does |
|---|---|
get_media_info |
Duration, resolution, codecs, fps, bitrate, file size |
trim_video |
Cut a clip by start time + end time/duration |
merge_videos |
Concatenate multiple clips into one (auto-normalizes resolution) |
convert_format |
Change container/codec (mp4, mov, webm, mp3, etc.) |
extract_audio |
Pull the audio track out as an MP3 |
resize_video |
Scale to a target width/height |
crop_video |
Crop to a rectangular region |
rotate_video |
Rotate 90/180/270° and/or flip horizontally/vertically |
change_speed |
Speed up or slow down (pitch-corrected audio) |
adjust_volume |
Raise/lower volume, or mute |
extract_frame |
Grab a still frame at a timestamp |
add_text_overlay |
Burn in a caption/title, optionally timed |
add_watermark |
Overlay a logo image at a chosen corner |
add_fade |
Fade in/out (video, audio, or both) |
compress_video |
Re-encode at a target quality (CRF) to shrink file size |
generate_gif |
Convert a video segment into an animated GIF |
Each tool returns the resulting file's path and size, so Claude can chain operations (e.g. trim → add watermark → compress) by feeding one tool's output into the next tool's input.
Testing without Claude
Run the unit tests, which generate a synthetic test video with ffmpeg and exercise every operation:
npm test
Run a full protocol-level test (spins up the server and a real MCP client, lists tools, and calls a couple of them):
node test/e2e-client.js
(The e2e test expects npm test to have been run first so test/tmp/v1.mp4
exists — or generate your own test clip there.)
Notes & limitations
- All operations re-encode by default for correctness (frame-accurate trims,
mixed-codec merges, etc.) rather than using fast
-c copycuts, so they're not the fastest possible, but they're reliable across arbitrary inputs. - This server runs entirely on your local machine and only touches files you point it at — no video data is uploaded anywhere.
- There's no built-in sandboxing beyond the
MEDIA_DIRconvenience for relative paths; since the server has the same filesystem permissions as whatever process launches it, only connect it to clients you trust.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。