A self-hosted archive that uses AI to organize your links, notes, and images
It is common to save interesting links or notes on a phone for later, only to forget their content over time. Karakeep collects this 'read later' data in one place and uses AI to automatically tag and summarize it, reducing the effort required to find items again. It is a self-hosted bookmark management tool that allows users to run their own server and control their data.
How data is stored
The tool manages links, text notes, images, and PDF files in a single list, going beyond a simple link repository. When saving a link, it automatically fetches the title, description, and preview image so you can immediately see what the content is. It also automatically collects content from RSS feeds or syncs with browser bookmarks to gather previously scattered materials in one place. It provides OCR functionality to extract text from images and an archiving feature that saves entire pages to prevent links from disappearing.
AI-based automatic organization and search
It uses LLMs to automatically tag and generate summaries for saved content. This process allows the use of local models via Ollama, making it suitable for users who prefer an environment where data does not leave their local setup. It supports full-text search and semantic search across all saved content, allowing you to easily find relevant materials without exact keywords. You can also configure a rule-based engine to automatically classify or manage bookmarks that meet specific conditions.
Access and collaboration across devices
In addition to the web app, it offers iOS and Android apps, allowing offline reading in mobile environments. Extensions for Chrome, Firefox, and Safari enable you to save bookmarks directly from the browser. It includes collaboration features that allow multiple users to manage the same list together, making it shareable for teams or families.
Self-hosting and cloud options
The project is designed with self-hosting as the priority. It is built on an open-source technology stack including Next.js, Drizzle, and Meilisearch, and can be easily deployed using Docker. For users who find running their own server burdensome, a paid managed service called Karakeep Cloud is available. It also supports importers for existing bookmark data from Pocket, Linkwarden, and Omnivore, simplifying the migration process.
Things to check before use
The tool is distributed under the AGPL-3.0 license and is owned by Localhost Labs Ltd. Using AI features requires access to an LLM model, and if using local models, you must consider hardware specifications. If you choose self-hosting, the responsibility for server maintenance lies with the user, making it more suitable for users with a technical background. The demo site is provided as read-only, so it is recommended to install it yourself or try the cloud version to verify actual functionality.
