immich-photo-manager

immich-photo-manager

Manage your self-hosted Immich photo library through conversation — natural language search via CLIP, geographic album curation, duplicate detection with perceptual hashing,

Category
访问服务器

README

<p align="center"> <img src="assets/icon.png" alt="immich-photo-manager" width="100"> </p>

<h1 align="center">immich-photo-manager</h1>

<p align="center"> <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a> </p>

📸🧹🗺️ MCP server for intelligent photo management with Immich — your self-hosted library, understood.

If you self-host Immich and your library has grown past the point where you can manage it by hand, immich-photo-manager gives Claude direct access to your Immich instance through 21 MCP tools and 11 specialized skills — from finding cross-ecosystem duplicates with perceptual hashing to generating interactive HTML galleries with a Cowork Actions Panel.

<p align="center"><img src="./assets/demo.gif" alt="immich-photo-manager demo" width="800"></p>


🚀 Quick Start

Prerequisites

Option A: Install as Claude Plugin (recommended)

git clone https://github.com/drolosoft/immich-photo-manager.git
cd immich-photo-manager

# Register the marketplace and install
claude plugin marketplace add .
claude plugin install immich-photo-manager

Verify with claude plugin list.

That's it. Open Claude and say "how healthy is my photo library?" to get started.

Note: The plugin connects to your Immich MCP server at localhost:8626. Make sure the server is running before using the plugin. See Option B for how to build and start the server.

Option B: Manual MCP Server Setup

If you prefer to configure the MCP server manually instead of using the setup script:

git clone https://github.com/drolosoft/immich-photo-manager.git
cd immich-photo-manager
pip3 install -r src/requirements.txt

Create .mcp.json in the project root:

{
  "mcpServers": {
    "immich": {
      "command": "python3",
      "args": ["-m", "immich_mcp_server"],
      "env": {
        "PYTHONPATH": "/ABSOLUTE/PATH/TO/immich-photo-manager/src",
        "MCP_TRANSPORT": "stdio",
        "IMMICH_BASE_URL": "https://your-immich-server.com",
        "IMMICH_API_KEY": "your-api-key-here"
      }
    }
  }
}

Important: Use PYTHONPATH pointing to the src/ directory. Do NOT use cwd (it is silently ignored by Claude Code).

Note on Smart Search (CLIP): The search_smart tool requires the Immich machine learning service to be running and Smart Search enabled in Administration > Settings > Machine Learning Settings > Smart Search. If the ML service is not configured, the tool will return a helpful error message instead of failing silently. All other tools work without the ML service. See the Immich Smart Search docs for setup details.


🎬 Core Workflow

You: "Create albums for all the places I've traveled"

→ Scans GPS data from 28,000+ photos
→ Clusters by location, identifies 47 destinations across 14 countries
→ Proposes album list for approval
→ Creates albums with curated selections (20-50 photos each)
→ Publishes to Gallery with shared links

✅ Created: 🇮🇹 Roma, Italia (47 photos, Jun 2023)
✅ Created: 🇪🇬 Cairo & Luxor, Egypt (63 photos, Mar 2024)
✅ Created: 🇲🇽 Oaxaca, México (38 photos, Dec 2022)
... 44 more albums

Geographic album creation combines GPS coordinates, CLIP visual search, and temporal matching — then filters out screenshots and duplicates automatically. One request, dozens of curated albums.


✨ Features

Feature What it does
🔍 AI-powered search Natural language photo search via CLIP ("sunset at the beach", "birthday cake")
🗺️ Geographic albums Create albums organized by place — GPS + CLIP combined for smart curation
🧹 Library cleanup Detect screenshots, duplicates, and low-quality images with multi-signal analysis
🔎 Duplicate report Deep cross-source duplicate analysis using perceptual hashing — finds re-encoded copies across Apple Photos, Google Photos, and other imports
🏥 Library health Comprehensive health check — asset inventory, metadata quality, storage breakdown, and recommendations
📅 Timeline gaps Find missing months, sparse periods, and single-source coverage risks in your photo timeline
🔧 Metadata fixer Detect and repair broken dates (noon/midnight), missing GPS, wrong timezones — with neighbor interpolation
📌 Auto-album curator Finds new photos that belong in existing albums using GPS, CLIP, and temporal matching
💾 Storage optimizer Identify RAW+JPEG pairs, oversized videos, and other space hogs with reclaimable space estimates
👥 People report Face recognition insights — who appears most, unnamed clusters worth naming, co-occurrence patterns
🌍 Travel map Interactive Leaflet.js map with clustered pins showing every place you've photographed
🔗 Gallery publishing Create shared links to make albums publicly accessible
🖼️ Interactive HTML galleries Self-contained gallery pages with embedded thumbnails, 3 themes (light/system/dark), 4 view modes, slideshow, and keyboard navigation
🎛️ Cowork Actions Panel Select photos in the gallery and copy batch commands (Create Album, Get EXIF, Find Similar, Download, etc.) straight into Cowork chat
🛡️ Safety first Never deletes automatically — always shows findings and asks before acting

🖼️ Interactive Gallery & Cowork Actions

Say "show me photos from Barcelona" and the plugin generates a self-contained HTML gallery — no server required, opens in any browser.

You: "Show me my Egypt album"

→ Fetches album + base64 thumbnails from Immich
→ Generates single-file HTML with embedded images
→ 3 themes (light/system/dark), 4 view modes, slideshow, keyboard nav
→ Cowork Actions Panel for batch operations

✅ egypt-album.html — open it, browse your photos, select & act

The Cowork Actions Panel is a sticky toolbar inside each gallery. Select photos, then click any action — the command is copied to your clipboard, ready to paste into Cowork chat:

Action What it copies
📋 Copy IDs Raw asset IDs for scripting
➕ Create Album "Create a new album with these photos: ..."
📂 Add to Album "Add these photos to album [name]: ..."
📦 Move to Album "Move these photos to a different album: ..."
🔍 Get EXIF Info "Get EXIF metadata for these photos: ..."
🔎 Find Similar "Find visually similar photos to: ..."
🖼️ Set as Cover "Set this photo as the album cover: ..."
⬇️ Download "Download these photos: ..."
❤️ Add to Favs "Mark these photos as favorites: ..."
⚠️ Remove from Album "Remove these from album [name]: ..."
🗑️ Delete Photos "Delete these photos permanently: ..."

The gallery is fully responsive, works offline once generated, and supports touch gestures on mobile.


🤔 Why immich-photo-manager?

Immich is excellent at storing, viewing, and searching your photos. But managing a large library — deduplication, metadata repair, album curation, storage analysis — still requires manual effort or custom scripts. immich-photo-manager bridges that gap.

Manual / scripts immich-photo-manager
🔍 Write API calls, parse JSON Natural language — "find my sunset photos from Italy"
🗺️ Export GPS, cluster manually Geographic albums — automatic GPS + CLIP + temporal matching
🧹 Hash files, diff checksums Perceptual hashing — finds re-encoded duplicates across import sources
📊 Query database, build reports Library health — one command for metadata quality, storage, recommendations
📅 SQL queries on timestamps Timeline gaps — detects empty months and single-source coverage risks
🔧 EXIF tools, manual review Metadata fixer — neighbor interpolation for missing GPS, broken timestamps
🛡️ Hope you don't delete the wrong thing Safety first — never deletes without showing findings and asking

🧩 Skills

Skills are specialized workflows that combine MCP tools with domain knowledge. Each handles a specific photo management task end-to-end.

Skill What it does
🗺️ Album Manager Create and curate albums organized by geography — GPS + CLIP + filtering
🔍 Photo Search Natural language search combining GPS, CLIP, and metadata filters
🧹 Photo Cleanup Detect screenshots, duplicates, and low-quality images with confidence levels
🔎 Duplicate Report Cross-source duplicate analysis using perceptual hashing (pHash)
🏥 Library Health Asset inventory, metadata completeness, storage breakdown, recommendations
📅 Timeline Gaps Detect empty months, sparse periods, single-source coverage risks
🔧 Metadata Fixer Repair broken dates, missing GPS, wrong timezones with neighbor interpolation
📌 Auto-Album Curator Find new photos that match existing albums using GPS, CLIP, and temporal patterns
💾 Storage Optimizer Identify RAW+JPEG pairs, oversized videos, reclaimable space estimates
👥 People Report Face recognition insights — top faces, unnamed clusters, co-occurrence patterns
🌍 Travel Map Interactive Leaflet.js map with clustered pins and heatmap overlay

See the Skills Reference for detailed workflows, triggers, and output formats.


⚙️ How It Works

Claude ←→ MCP (stdio) ←→ Python Server ←→ Immich REST API
                                              your-instance

The MCP server is a Python module (immich_mcp_server) that communicates with Claude over stdio transport. It exposes 21 tools that map to Immich REST API endpoints, with added intelligence for batch operations, thumbnail embedding, and credential management.

Gallery HTML files are fully self-contained — thumbnails are fetched from Immich, encoded as base64, and embedded directly in the HTML. This is required because the Cowork viewer runs in an about: sandbox that blocks all external network requests. See Architecture for the full technical explanation.

21 MCP Tools

Category Tools
🏥 Health (3) ping, get_server_version, get_statistics
📷 Assets (2) get_asset_info, get_map_markers
🔍 Search (2) search_metadata, search_smart (CLIP)
📁 Albums (7) list_albums, get_album, create_album, update_album, delete_album, add_assets_to_album, remove_assets_from_album
🔗 Sharing (2) list_shared_links, create_shared_link
🖼️ Thumbnails (3) get_asset_thumbnail, get_album_thumbnails, get_thumbnails_batch
🔧 Config (2) get_connection_info, update_credentials

See the MCP Tools Reference for parameters, return types, and examples.


📚 Documentation

Document Description
Getting Started Installation, configuration, deployment options, and troubleshooting
Architecture How base64-embedded thumbnails solve the Cowork sandbox restriction
Skills Reference All 11 skills — workflows, triggers, parameters, output formats
MCP Tools Reference All 21 MCP tools — parameters, return types, examples
CORS Setup Guide Optional — enable direct URL thumbnail loading for browser-viewed galleries

Additional dependencies (optional)

Some advanced skills require Python packages or direct database access:

pip3 install Pillow imagehash pillow-heif
Package Used by Purpose
Pillow duplicate-report Image loading
imagehash duplicate-report Perceptual hashing (pHash)
pillow-heif duplicate-report HEIC/HEIF support (Apple Photos)
PostgreSQL client library-health, timeline-gaps, people-report, storage-optimizer Database-level analysis

🤝 Contributing

Contributions are welcome. If you find a bug, have a feature idea, or want to improve a skill:

  1. Fork the repository
  2. Create a feature branch (git checkout -b feature/my-feature)
  3. Commit your changes
  4. Open a pull request

For bug reports and feature requests, use GitHub Issues.


☕ Support

If immich-photo-manager saved you time or made your photo library easier to manage, consider buying me a coffee — it keeps the next one coming!

<p align="center"> <a href="https://buymeacoffee.com/juan.andres.morenorub.io"><img src="https://img.shields.io/badge/Buy%20Me%20A%20Coffee-ffdd00?style=for-the-badge&logo=buy-me-a-coffee&logoColor=black" alt="Buy Me A Coffee"></a> </p>


📜 License

MIT License — free to use, modify, and distribute.

Forged by Drolosoft · Tools we wish existed

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选