Self-Host Hoarder: AI-Powered Bookmark Manager 2026

TL;DR

Hoarder (AGPL 3.0, ~10K GitHub stars, Next.js) is a self-hosted bookmark manager with AI superpowers. Save a URL and Hoarder automatically fetches the full page, takes a screenshot, extracts text, and uses a local AI (Ollama) to generate tags — zero manual organization required. Raindrop.io Pro ($3/month) offers similar AI features but sends your data to their cloud. Hoarder runs locally with complete privacy.

Key Takeaways

• Hoarder: AGPL 3.0, ~10K stars, Next.js — AI-first bookmark manager with automatic tagging
• Ollama integration: Uses local LLMs (llama3.2, mistral, etc.) for private AI tagging
• Full-page archiving: Saves snapshots of pages so they're available even if the original goes down
• Screenshots: Visual thumbnails for every bookmark — makes browsing bookmarks fast
• Browser extensions: Chrome and Firefox — one-click saving with instant AI processing
• Mobile apps: iOS and Android — save links from any app via share sheet

---

Hoarder vs Linkding vs Raindrop.io

Feature	Hoarder	Linkding	Raindrop.io Pro
AI auto-tagging	Yes (Ollama)	No	Yes (cloud)
Full-page archive	Yes	No	Yes
Screenshots	Yes	No	Yes
Privacy	Local	Local	Cloud
Setup complexity	Medium	Simple	None
Mobile apps	Yes	PWA	Yes
Resource usage	~500MB+	~50MB	N/A
Price	Free	Free	$3/mo

---

Part 1: Docker Setup

Hoarder requires three services: the app itself, MeiliSearch (fast full-text search), and Chromium (for page archiving and screenshots).

# docker-compose.yml
services:
  web:
    image: ghcr.io/hoarder-app/hoarder:latest
    container_name: hoarder
    restart: unless-stopped
    ports:
      - "3000:3000"
    volumes:
      - hoarder_data:/data
    environment:
      # Required:
      NEXTAUTH_SECRET: "${NEXTAUTH_SECRET}"    # openssl rand -base64 36
      NEXTAUTH_URL: "https://hoarder.yourdomain.com"

      # MeiliSearch:
      MEILI_ADDR: "http://meilisearch:7700"
      MEILI_MASTER_KEY: "${MEILI_MASTER_KEY}"  # openssl rand -base64 24

      # Chromium for archiving:
      BROWSER_WEB_URL: "http://chrome:9222"

      # Data directory:
      DATA_DIR: /data

      # Optional: Ollama for AI tagging:
      OLLAMA_BASE_URL: "http://ollama:11434"
      OLLAMA_MODEL: "llama3.2"

      # Optional: disable signup after first user:
      # DISABLE_SIGNUPS: "true"

    depends_on:
      - meilisearch
      - chrome

  meilisearch:
    image: getmeili/meilisearch:v1.6
    container_name: hoarder-meilisearch
    restart: unless-stopped
    volumes:
      - meilisearch_data:/meili_data
    environment:
      MEILI_MASTER_KEY: "${MEILI_MASTER_KEY}"
      MEILI_NO_ANALYTICS: "true"

  chrome:
    image: gcr.io/zenika-hub/alpine-chrome:latest
    container_name: hoarder-chrome
    restart: unless-stopped
    command: >
      --no-sandbox
      --disable-gpu
      --disable-dev-shm-usage
      --remote-debugging-address=0.0.0.0
      --remote-debugging-port=9222
      --hide-scrollbars

  # Optional: local AI with Ollama
  ollama:
    image: ollama/ollama:latest
    container_name: hoarder-ollama
    restart: unless-stopped
    volumes:
      - ollama_data:/root/.ollama
    # Uncomment for GPU:
    # deploy:
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: 1
    #           capabilities: [gpu]

volumes:
  hoarder_data:
  meilisearch_data:
  ollama_data:

# Generate secrets:
echo "NEXTAUTH_SECRET=$(openssl rand -base64 36)" >> .env
echo "MEILI_MASTER_KEY=$(openssl rand -base64 24)" >> .env

docker compose up -d

# Pull the AI model (after Ollama starts):
docker exec hoarder-ollama ollama pull llama3.2

---

Part 2: HTTPS with Caddy

hoarder.yourdomain.com {
    reverse_proxy localhost:3000
}

Visit https://hoarder.yourdomain.com → create your account.

---

Part 3: Browser Extensions

Chrome

1. Install Hoarder Extension
2. Extension → Options:
   • Server URL: https://hoarder.yourdomain.com
   • Click Login → authenticates via your account
3. Click extension icon on any page → Save

Firefox

1. Install from Mozilla Add-ons: search "Hoarder"
2. Same configuration

When you save a link:

1. Hoarder immediately stores the URL
2. Chromium fetches the full page + screenshot (background)
3. Ollama generates tags from the content (background)
4. Full text becomes searchable

---

Part 4: Mobile Apps

iOS

1. Install Hoarder for iOS
2. Server: https://hoarder.yourdomain.com
3. Login with your credentials
4. Share Sheet → Hoarder — save any link from Safari, Twitter, etc.

Android

1. Install from Play Store: search "Hoarder"
2. Same setup
3. Share links from any app directly to Hoarder

---

Part 5: AI Tagging Configuration

Using Ollama (local, private)

# In docker-compose.yml:
environment:
  OLLAMA_BASE_URL: "http://ollama:11434"
  OLLAMA_MODEL: "llama3.2"    # or mistral, phi3, etc.

# Available models for tagging (CPU-friendly):
docker exec hoarder-ollama ollama pull llama3.2       # 2GB, good quality
docker exec hoarder-ollama ollama pull phi3:mini       # 2.3GB, fast
docker exec hoarder-ollama ollama pull mistral:7b      # 4.1GB, best quality

# Switch model without restart:
# Update OLLAMA_MODEL env var and restart hoarder container
docker compose restart web

Using OpenAI API (if you prefer cloud)

environment:
  OPENAI_API_KEY: "sk-..."
  OPENAI_BASE_URL: "https://api.openai.com/v1"    # or any OpenAI-compatible API
  INFERENCE_TEXT_MODEL: "gpt-4o-mini"
  INFERENCE_IMAGE_MODEL: "gpt-4o-mini"

Prompt customization

Hoarder uses a default system prompt for tag generation. Override in settings:

# Settings → AI → Custom Prompt
You are a bookmark tagging assistant. Generate 3-5 short, lowercase tags
for the provided content. Focus on: topic, type (article/tool/video/paper),
and primary programming language or technology stack.
Return only comma-separated tags.

---

Part 6: Lists and Organization

Lists (manual organization)

Beyond AI tags, organize bookmarks into lists:

1. Lists → + New List: Reading Queue, Work Research, Tools to Try
2. Drag bookmarks into lists, or save directly to a list from the extension
3. Lists are shareable — each gets a public URL

Filtering

# In the search bar:
#docker              — filter by tag
list:work            — filter by list
is:archived          — show archived
is:unread            — show unread

# Combine:
#kubernetes is:unread

---

Part 7: REST API

# Get API key:
# Settings → API Keys → Generate

API_KEY="your-api-key"
BASE="https://hoarder.yourdomain.com"

# List bookmarks:
curl "$BASE/api/v1/bookmarks" \
  -H "Authorization: Bearer $API_KEY" | jq '.bookmarks[].title'

# Add a bookmark:
curl -X POST "$BASE/api/v1/bookmarks" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "link",
    "url": "https://example.com",
    "title": "Optional override title"
  }'

# Get bookmark with AI tags:
curl "$BASE/api/v1/bookmarks/BOOKMARK_ID" \
  -H "Authorization: Bearer $API_KEY" | jq '{title: .title, tags: .tags}'

# Archive a bookmark:
curl -X PUT "$BASE/api/v1/bookmarks/BOOKMARK_ID" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"archived": true}'

---

Part 8: Import from Other Services

Import from Pocket

1. Pocket → Export → downloads ril_export.html
2. Hoarder → Settings → Import → select Pocket HTML file
3. All links import with original tags preserved

Import from Raindrop.io

1. Raindrop.io → Export → CSV or HTML
2. Hoarder → Settings → Import → select file

Import bookmarks from browser

1. Chrome/Firefox: Bookmarks Manager → Export → HTML
2. Hoarder → Settings → Import → select HTML file

All imported bookmarks are queued for AI processing automatically.

---

Resource Requirements

Component	Minimum RAM	Recommended RAM
Hoarder web	256MB	512MB
MeiliSearch	256MB	512MB
Chromium	512MB	1GB
Ollama (llama3.2)	4GB	8GB
Total	~1GB (no Ollama)	~6-10GB (with Ollama)

If RAM is limited, use OpenAI API instead of Ollama — it requires no local inference.

---

Maintenance

# Update:
docker compose pull
docker compose up -d

# Backup:
tar -czf hoarder-backup-$(date +%Y%m%d).tar.gz \
  $(docker volume inspect hoarder_hoarder_data --format '{{.Mountpoint}}')

# Backup MeiliSearch indexes:
tar -czf hoarder-search-$(date +%Y%m%d).tar.gz \
  $(docker volume inspect hoarder_meilisearch_data --format '{{.Mountpoint}}')

# Trigger re-processing of all bookmarks (after AI model change):
# Settings → Re-run AI tagging on all bookmarks

# Logs:
docker compose logs -f web
docker compose logs -f chrome

---

See also: Linkding — simpler, minimal bookmark manager with lower resource requirements

See all open source productivity tools at OSSAlt.com/categories/productivity.