Karakeep
Tabs
Karakeep: The Bookmark Everything App — and Why You Should Self-Host It
Save links, notes, images, and PDFs — and let AI do the organizing for you.
There's a certain kind of person who opens twenty browser tabs "just to come back to later," pastes links into Slack messages to themselves, and has a Notes app full of half-finished ideas. If that sounds familiar, you've probably tried — and abandoned — more than a few bookmark managers.
Karakeep is the one that might finally stick. It's open-source, AI-powered, beautifully cross-platform, and — crucially for the self-hosting crowd — trivially easy to deploy on your own hardware.

What Is Karakeep?
Karakeep (formerly known as Hoarder) bills itself as "The Bookmark Everything App," and it earns that title. At its core, it's a self-hosted bookmark and content manager that goes well beyond just saving URLs. You can save:
- Links — web pages, articles, YouTube videos, documentation
- Notes — quick text snippets, ideas, or reference content
- Images — screenshots, photos, visual references
- PDFs — documents you want to keep indexed and searchable
When you save something, Karakeep automatically fetches the title, description, and a preview image. More importantly, it uses AI to automatically tag your content — so the moment you save an article about OPNsense firewall rules, it's already tagged with networking, firewall, and opnsense before you've done a thing.
The project has earned over 24,000 stars on GitHub and has attracted more than 150 contributors, making it one of the more active open-source personal productivity tools in the homelab space.
Features Worth Knowing About
AI-Powered Tagging
This is the killer feature. Rather than forcing you to manually categorize every bookmark (which is why most bookmark managers eventually get abandoned), Karakeep analyzes your saved content and assigns tags automatically. You can use OpenAI's API for this, or — for those who want to keep everything local — Ollama for fully offline AI inference. Yes, your homelab's Ollama instance can power this.
Full-Text Search
All your saved content is indexed and fully searchable using Meilisearch under the hood. Search doesn't just match titles — it reaches into the full content of saved pages.
RSS Feed Hoarding
Connect RSS feeds and Karakeep will automatically pull in new items as they're published. If you follow a dozen tech blogs, Karakeep becomes your own private, self-hosted RSS reader and archive in one.
Highlights
Read an article and want to save a specific passage? Karakeep lets you highlight text on any saved page and keeps those highlights organized separately for quick reference.
Rule Engine
Power users will appreciate the automation rules. You can set up rules to automatically tag, organize, or route incoming bookmarks based on URL patterns, content, or source — think of it as a lightweight filter system for your personal knowledge base.
Collaborative Lists
Share curated lists with others. Useful if you're building a shared resource library with a team or household.
API & Webhooks
Karakeep exposes a full REST API and supports webhooks, making it straightforward to integrate with other tools — automation pipelines, n8n, custom scripts, you name it.
Cross-Platform Apps and Extensions
This is where Karakeep really distinguishes itself from other self-hosted options. It's not just a web app:
- iOS and Android native apps
- Chrome extension
- Firefox add-on
- Safari extension
- CLI tool
You can capture from anywhere and it goes straight to your self-hosted instance.
Why Self-Host It?
The cloud-hosted version of Karakeep exists (and is now in public beta), but self-hosting is where the project shines — and where it makes the most sense for homelab users.
Your data stays yours. Every bookmark, note, image, and PDF is stored on hardware you control. There's no third-party indexing your reading habits or surfacing your saved content to advertisers.
Local AI inference. Pointing Karakeep at a local Ollama instance means the AI tagging pipeline never leaves your network. For anyone running a GPU-backed or even CPU-based Ollama setup, this is a compelling zero-cloud-cost option.
No subscription. Run it on a spare Raspberry Pi and it costs you nothing beyond electricity.
Custom integrations. With full API access and your own deployment, you can wire Karakeep into any part of your homelab workflow.
Self-Hosting on a Raspberry Pi 5 with Portainer
Karakeep is a multi-container application. The stack includes:
- karakeep-web — the main application (Next.js)
- karakeep-chrome — a headless Chromium instance for full-page screenshots and archival
- meilisearch — the full-text search engine
The official documentation points to a standard Docker Compose setup, and it maps directly to a Portainer Stack — which is exactly what we're using here on a Raspberry Pi 5.
Why Portainer?
Portainer gives you a browser-based GUI to manage your Docker stacks, watch container logs, and handle updates — all without SSH-ing into your Pi every time. If you're already running Portainer in your homelab, Karakeep drops right in.
Prerequisites
Before deploying, you'll need:
- Docker and Docker Compose installed on your Pi (or Portainer CE already running)
- A few minutes to generate two random secrets
- Optionally: an OpenAI API key or a local Ollama endpoint for AI tagging
Generate your secrets with:
openssl rand -base64 36
Run this twice — once for NEXTAUTH_SECRET, once for MEILI_MASTER_KEY.
Portainer Stack YAML
Drop this into Portainer under Stacks → Add Stack → Web Editor. Update the environment variables with your own values before deploying.
Environment Variables: The Key Ones
VariableWhat It DoesNEXTAUTH_SECRETSigns authentication tokens — keep this secret and randomMEILI_MASTER_KEYSecures your Meilisearch instanceNEXTAUTH_URLThe URL where you access Karakeep (e.g., http://karakeep.home.arpa)KARAKEEP_VERSIONUse release for latest stable, or pin a version like 0.32.0OPENAI_API_KEYOptional: enables cloud AI tagging via OpenAIOLLAMA_BASE_URLOptional: point to your local Ollama for private AI taggingDISABLE_SIGNUPSSet to true after creating your account to lock registration
Connecting AI: OpenAI vs. Ollama
Karakeep supports multiple AI backends for its tagging engine.
OpenAI is the simplest to set up — add your API key and it just works. The cost is minimal for personal use; the Karakeep docs note that tagging a typical personal bookmark collection costs only a few cents per month.
Ollama is the self-hosters' choice. If you're already running Ollama on your network (say, on a dedicated machine with a GPU, or even on the Pi itself for lighter models), you can point Karakeep at it:
OLLAMA_BASE_URL=http://10.11.12.22:11434OLLAMA_MODEL=llama3
The trade-off is that local models may be less accurate at tagging than GPT-4o-mini, but for most personal content the results are perfectly usable — and nothing leaves your network.