feather
MCP server for Feather that lets AI agents compress videos, images, GIFs, and PDFs locally via CLI tools for compression, probing, estimation, and history.
README
Feather
Make videos, images, GIFs and PDFs lighter — entirely on your device.
Feather is a fast, open-source media compressor for macOS (Windows & Linux builds coming). Drop files in, pick a quality, get smaller files. Nothing is uploaded, ever.
<p align="center"> <video src="public/feather-trailer.mp4" width="820" autoplay loop muted playsinline></video> </p> <p align="center"><a href="public/feather-trailer.mp4">▶ Watch the trailer</a></p> <p align="center"> <a href="https://github.com/FirasLatrech/feather/releases/latest">Download</a> · <a href="#cli--mcp-ai-agents">CLI & MCP</a> </p> <p align="center"><i>Video · Image · GIF · PDF — hardware-accelerated, batch, folder watching, CLI + MCP</i></p>
Download
Easiest — one line in Terminal (downloads the latest release, installs to Applications, and grants it access to open):
curl -fsSL https://raw.githubusercontent.com/FirasLatrech/feather/main/install.sh | sh
Or grab the .dmg from the latest release →.
Feather isn't notarized yet (it's open source, not signed by a paid Apple developer account), so if you install the
dmg by hand macOS says "Feather is damaged" — it isn't; run xattr -cr /Applications/Feather.app once and open it.
FFmpeg is downloaded by Feather itself on first launch if you don't have it. Ghostscript (PDF only) can be installed from Settings.
Features
- Video — H.264 / H.265 / VP9 / AV1, hardware encoding (Apple VideoToolbox) always on, target file size, resize, frame rate, trim, remove audio, video → GIF, extract MP3. Bitrate-aware: never spends more bits than the source.
- Images — JPG (mozjpeg), PNG (oxipng), WebP, AVIF, HEIC/TIFF input, resize, keep metadata. Never produces a larger file.
- GIF — per-file palette + dithering, resize, fps.
- PDF — Ghostscript presets from print quality to screen.
- Batch with bounded parallelism, live speed × / ETA, size & time estimates before you press Compress.
- Auto-compress: watch folders (e.g. Downloads) and compress whatever lands there; optionally replace the original.
- Per-file setting overrides, output folder / naming templates, replace-in-place, move to Trash, keep dates.
- History with stats, notifications, light/dark theme.
CLI & MCP (AI agents)
Feather ships feather-cli inside the app bundle (Feather.app/Contents/MacOS/feather-cli) — a command-line
tool and an MCP server, so Claude, Cursor or any MCP client can compress
files on your machine.
# Terminal
feather-cli compress ~/Movies/*.mp4 --quality good --max 1920
feather-cli compress photo.heic --format webp --out ~/Desktop
feather-cli probe video.mov · feather-cli estimate *.png --quality medium · feather-cli history
# Claude Code
claude mcp add feather -- /Applications/Feather.app/Contents/MacOS/feather-cli mcp
# Claude Desktop / Cursor (mcp.json)
{ "mcpServers": { "feather": { "command": "/Applications/Feather.app/Contents/MacOS/feather-cli", "args": ["mcp"] } } }
MCP tools: compress (paths, quality, format, max, out, replace, no_audio, codec, target_mb — with progress
notifications), probe, estimate, history. Destructive options (replace) are only applied when passed explicitly.
Settings → AI agents · MCP shows the exact commands with copy buttons.
Requirements
- macOS 13+ (Apple Silicon or Intel)
- FFmpeg —
brew install ffmpeg - Ghostscript for PDFs (optional) —
brew install ghostscript
Development
pnpm install
pnpm tauri dev # desktop app
pnpm dev # UI only, in the browser with a mocked backend (add ?demo=files)
cd src-tauri && FEATHER_SAMPLES=/path/to/samples cargo test # engine tests against real ffmpeg/gs
Stack: Tauri 2 (Rust) · React · TypeScript · Vite.
Engine: FFmpeg for video/GIF, Ghostscript for PDF, pure-Rust image codecs (image, mozjpeg, oxipng, webp, ravif).
Screenshots
| Drop zone | Dark theme, batch running |
|---|---|
![]() |
![]() |
Contributing
Issues and PRs are welcome. Run pnpm tauri dev for the app, pnpm dev for the UI in a browser (mock backend,
?demo=files|running|done), and cargo test in src-tauri (set FEATHER_SAMPLES to a folder with sample media
to run the encoder tests).
Roadmap
See PLAN.md.
License
MIT © Firas Latrach
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。

