Ouestcharlie Woof

io.github.ouestcharlie由 GitHub 账号 ouestcharlie 发布v1.0.0更新于 Oct 25, 2026

An MCP server and app to search and browse your photos while preserving your privacy

已验证STDIO仅桌面文件与存储媒体与设计

概览

AI 生成的概览

一个本地 MCP 应用,让 AI 助手成为你自己照片和视频库的可搜索浏览器,文件始终留在你的硬盘上。

可以做什么
以网格或预览方式浏览本地照片和视频库,缩略图和详情由本机提供,而非经由助手。按描述文本、标签、评分、日期和分区搜索,支持与 darktable 和 Lightroom 共享的层级标签。按日期、尺寸和 GPS 边界框筛选照片和视频,按品牌、型号、光圈和镜头筛选相机。可按任意字段升序或降序排序。检测新增或删除的文件以及其他工具编辑过的附属文件以重新索引。
适用场景
适合拥有大量本地照片或视频收藏,并希望在桌面 AI 助手中通过描述、标签、日期或相机信息查找图片,而不把原图上传到云服务的场景。
运行要求
以本地 Python 包(PyPI)通过 stdio 运行,需要支持 MCP App 的桌面应用,例如 Claude Desktop、Goose 或 VS Code。支持 macOS、Windows 和 Linux。照片库必须位于本地或已同步到挂载的驱动器。未声明需要账户、API 密钥或环境变量。
安装前请注意
搜索时只有元数据(EXIF、GPS 坐标、相机信息、日期、文件路径)会发送给 AI 助手;除非你明确分享,原图不会上传。Woof 以只读方式读取照片库,不会修改、移动或删除原始文件,但会在照片旁写入 XMP 附属文件和 .ouestcharlie 文件夹。目前为早期预览版,自动重新索引仅部分实现。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Ouestcharlie Woof,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

[Woof]

Woof — See your Photo Gallery in your AI assistant

[Status: Early Preview]

[macOS] [Linux] [Windows]

[MCP Server] [uv]

mcp-name: io.github.ouestcharlie/ouestcharlie-woof

Woof is the photo and video gallery companion to your your AI assistant (Claude Desktop, Goose, VS Code / GitHub Copilot...). It complements those powerful tools with a searchable gallery. Your photos and videos remain exactly where they are — on your own drives (local or mounted).

No cloud subscription. No proprietary lock-in. Your library, your way.

Woof is the MCP App frontend to "Où est Charlie ?" ("Where is Wally?" in French), a full AI native framework to manage your photos and videos.

What makes it different

Most photo managers lock your library into a cloud service (Google Photos, iCloud) or require a database server that becomes a single point of failure. Woof takes a different approach:

  • Conversation as your gallery. Woof connects to your AI assistant (Claude Desktop, Goose, VS Code / GitHub Copilot…) and turns it into a full photo browser. Ask in plain language, get results inline. No separate app to learn.
  • Privacy by design. Only metadata travels to your AI assistant — your actual photos are served locally by Woof. Your pictures are never uploaded to any AI service unless you explicitly ask.
  • No database lock-in. Metadata lives as XMP sidecar files right next to your photos, plus lightweight JSON manifests. Move a drive, copy a folder — your entire organization travels with your photos.
  • Open formats, forever. XMP is an ISO standard. JSON is universal. AVIF is royalty-free. Every tool you already use — Lightroom, darktable, ExifTool — can read your metadata today and long after OuEstCharlie is gone.
  • Your photos are never touched. Woof reads your library as-is. It never modifies, moves, or deletes your original files. It also honors existing XMP metadata from Lightroom, darktable, or any other tool — rather than overwriting it.
  • Works with your existing folder structure. Just point Woof at your photos folder. No migration, no reorganization required.

More about OuEstCharlie and Woof on the OuEstCharlie Blog


Installation and first steps

See OuEstCharlie Woof install and first steps


Status

Woof is an early preview. It works well today for browsing and searching a local photo library.

Current features

FeatureNotes
Compatible with Claude Desktop, Goose, VS CodeAny desktop application with MCP App support
Available on Macos, Windows, LinuxPython packages are built and released in Pypi
Local filesystem indexingFiles must be locally synced
Photos (JPEG, PNG, TIFF, HEIC, RAW)HEIC and RAW depend on the build options
Video support (MOV, MP4)
Gallery view as grid or previewPhoto details on preview
Image thumbnails and previewsOptimized for display
Search description, tags, rating, date, partitionfull text search on description
Hierarchical tags (Places|Europe|France), shared with darktable and Lightroomsearch a whole branch or a tag at any level; darktable's own tags are left out
Search photo features (date, dimensions, GPS bounding box)
Search video features (duration, dimensions, GPS bounding box)
Search camera features (make, model, aperture, lens)
Sort ascending or descending on any field
Change detection / automatic re-indexingPartial — added and removed pictures, and sidecars edited by other tools (darktable, digiKam…); indexing is not yet triggered automatically

Planned features

Feature
Albums and smart filters
Share pictures with host (Claude Desktop, ChatGPT, Goose…)
Enrichment agents (faces, scene recognition)
Mobile companion app
Native cloud libraries (S3, OneDrive, GCS…)

If you hit a bug or unexpected behavior, please open an issue.

[Woof in VSCode]

Ouestcharlie Woof photo gallery inside VSCode, an alternative to Claude CoWork for techies


Privacy Policy

Woof is designed with privacy as a core principle.

  • Data collected: Only photo metadata (EXIF, GPS coordinates, camera make/model, dates, file paths) is read and indexed. No account or personal information is collected.
  • Data storage: All metadata is stored locally on your own device as XMP sidecar files and JSON manifests alongside your photos. No data is stored on any remote server.
  • AI assistant: Only metadata is sent to your AI assistant (Claude, ChatGPT, Goose…) when you perform a search. Thumbnails and previews remain in the MCP App gallery widgets: they are served by Woof from your machine (loopback only) and never pass through the assistant. Your original photo files are never uploaded to any AI service unless you explicitly share them.
  • Third parties: No metadata or usage data is shared with any third party.
  • Retention: All data remains under your full control. Deleting the XMP sidecars and .ouestcharlie/ folders from your photo library completely removes all Woof metadata.

For privacy questions, please open an issue.


Support

Bug reports and feature requests: GitHub Issues


Developers' corner

For developer and architecture documentation, see README_DEV.md.

### Contributing

See the Contributing section in OuEstCharlie


License

MIT license

来源:README.md,提交 14c270e

工具

0
SourceWeft 尚未在沙箱中启动过该软件包,因此其工具未知。

版本历史

1
  1. v1.0.0最新Sep 16, 2026

评分

已满足适用检查项的 50%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分

    请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分