Most bookmark tools are just lists of URLs. The URL points somewhere, the page might still exist, and there is no copy if it disappears. For casual saving that is fine. For anyone doing serious research, building a knowledge base, or relying on sources over months or years, it is a structural problem. Webpages move, sites shut down, and paywalls appear where free content used to live. Linkwarden solves this by turning each bookmark into an archive.
It is a self-hosted, open-source collaborative bookmark manager that automatically saves a screenshot, a PDF, and a single-file HTML copy of every webpage you bookmark, at the moment you bookmark it. The idea is that even if the original URL goes dead tomorrow, you still have the full content. The project has 19,000+ stars and released v2.16.0 in late July 2026. The source lives at github.com/linkwarden/linkwarden.
| Fact | Value |
|---|---|
| What it is | A self-hosted collaborative bookmark manager with full-page archiving and annotations |
| Built by | Daniel (daniel31x13) and open-source contributors |
| License | AGPL (open-source license that requires derivative works to also be open-source) |
| Platforms | Web, iOS, Android, PWA (a website installable as an app on your home screen); browser extensions for major browsers |
| Requires | Docker (software that packages the app into a container you can run on any server) and a server with 2+ GB RAM |
| Install method | Docker Compose (a tool for running multi-container Docker setups with one command) with a three-step setup |
| Verdict | The right choice for researchers and knowledge workers who want preserved, annotatable, searchable bookmarks on infrastructure they control |
What Linkwarden actually is
Linkwarden is what happens when you take browser bookmarks seriously. The core behavior is simple: when you save a link, Linkwarden immediately fetches the page and stores three copies, a screenshot image, a PDF, and a self-contained HTML file (a single-file snapshot of the page that includes all its images and styles). This process is called archiving ("saving a snapshot of the webpage at the time you bookmark it, so the content is preserved even if the original URL goes dead").
That archived copy lives on your server, under your control, for as long as you choose to keep it. No external dependency, no account at a third-party archive service, no link-rot risk.
On top of preservation, Linkwarden adds a reader view that strips ads and navigation from the saved content, leaving just the text and images. Inside that reader you can highlight passages and write annotations tied to specific sections. This is the feature that makes Linkwarden useful for active research rather than passive filing.
Organization runs through collections ("folders for organizing bookmarks in Linkwarden"), which can be nested into sub-collections. Each link can belong to a collection and carry multiple tags. Search works across full text, not just titles and URLs, because the content is stored locally.
The collaboration side is real and not an afterthought. You can invite other users to a collection and assign them specific permission levels: view, edit, or admin. This makes Linkwarden useful for a research team, a journalism shop, or any group that needs a shared reference library. Public sharing is available too; a collection can be published to a URL anyone can visit without logging in.
The AGPL license means the code is fully open. You can run it on your own hardware, inspect every line, and modify it. The cloud option at linkwarden.app is $3 per user per month ($2.25 with annual billing) and caps each user at 30,000 links. Self-hosted is uncapped.
What you get after installing it
The main interface organizes around a sidebar of your collections. Clicking a collection shows its links as cards with the archived screenshot as a thumbnail. Each card opens the full bookmark view: the original URL, the three preserved formats, the reader view with your annotations, and the tag list.
Archiving on save is automatic. Every new link triggers a background job that fetches and stores the screenshot, PDF, and HTML on your own hardware, with no external archiving service involved. Bulk import from browser bookmarks or from Raindrop, Pocket, and Instapaper queues everything for archiving.
Full-text search reaches into the archived HTML content, not just titles and URLs. The search supports advanced operators for filtering by collection, tag, and date.
AI tagging is an optional local feature. Connect an Ollama instance (a tool for running AI language models on your own machine) and Linkwarden will suggest tags based on the archived page content, entirely on your server.
Browser extension is open-source. Click it on any page, optionally add tags and a note, and the bookmark saves with archiving queued immediately.
RSS feed following lets you subscribe to a site's feed so new items arrive as bookmarks in a collection, archived automatically.
iOS and Android apps are available in the App Store and Google Play, with the July 2026 v2.16.0 release adding mobile support for re-generating preserved formats.
The install experience
The install requires Docker and assumes a Linux server with about 2 GB of RAM. Three commands fetch the configuration files:
mkdir linkwarden && cd linkwarden
curl -O https://raw.githubusercontent.com/linkwarden/linkwarden/refs/heads/main/docker-compose.yml
curl -L https://raw.githubusercontent.com/linkwarden/linkwarden/refs/heads/main/.env.sample -o ".env"
You then open the .env file and set three secrets: NEXTAUTH_SECRET (the authentication key), MEILI_MASTER_KEY (the key for Meilisearch, the search engine that indexes your bookmarks), and POSTGRES_PASSWORD (the password for the local database). PostgreSQL ("the database that stores your bookmarks locally") runs as one of the containers alongside Meilisearch, so no external database or search setup is required. Each secret needs to be unique, and ones with special characters should be wrapped in quotes.
Then docker compose up starts everything. The first run downloads and builds the containers, which takes a few minutes on a typical connection. After that, the app is available at http://localhost:3000 and you create the admin account in the browser.
The documentation at docs.linkwarden.app covers the full setup including reverse-proxy configuration (setting up a traffic router that forwards requests from your domain to the running container) for exposing the app at a custom domain with HTTPS.
Two gotchas worth knowing before you start. First, the initial page-preservation queue can be heavy if you import a large existing bookmark collection. A few hundred links queuing for archiving simultaneously will use noticeable CPU and network bandwidth, then settle down once the queue drains. Second, the archived files accumulate on disk over time. A collection of several thousand bookmarks with full screenshots and PDFs will use several gigabytes. Plan storage accordingly.
Where it fits and what to compare it to
Linkwarden is not a minimal bookmark tool and is not trying to be. If you want something that syncs browser bookmarks with tagging and a clean UI but you do not need archiving or annotations, Raindrop.io is the better fit. Raindrop is hosted, has a polished mobile experience, and offers a functional free tier. It does not archive pages by default.
Pocket is the dominant read-it-later tool and handles the "save for later" workflow well. It does not offer self-hosting and does not give you archived copies in multiple formats.
Wallabag is a closer self-hosted equivalent. It saves a readable copy of each article but does not capture a full-page screenshot or PDF, so it loses on archiving fidelity.
Karakeep (an open-source bookmark manager previously named Hoarder) is a newer project in the same category with AI tagging as a core feature. Worth watching, but it has less community history than Linkwarden's four-year run.
The decision comes down to archiving fidelity. If pages disappearing is the problem you are solving, Linkwarden's triple-format preservation (screenshot, PDF, HTML) is the most complete answer in the self-hosted category. If archiving is a secondary concern, a hosted tool like Raindrop carries less maintenance overhead.
Verdict
Linkwarden earns its place for a specific kind of user: the researcher, analyst, journalist, or knowledge worker who saves links to work with them, not just file them. The combination of automatic archiving in three formats, in-place annotation, full-text search, and collaborative collections covers the full workflow from capture to long-term retrieval.
The install is straightforward for anyone comfortable with Docker. The ongoing maintenance is minimal once it is running. The AGPL license and self-hosted model mean the data is yours, the link limit is yours to set, and the tool keeps working regardless of what happens to the Linkwarden company or cloud offering.
The tradeoff is infrastructure ownership. You are responsible for backups, uptime, and disk space. If a managed service is the right fit for your situation, the cloud tier at $3 per user per month is a reasonable price for what it provides, and the data export option means you are not locked in.
For anyone who has lost an important source to link rot and decided to take the problem seriously, this is the tool to install.