Slash: a link shortener you host, where the whole app is one binary and one port
An open source, self-hosted platform for sharing and managing your most frequently used links. Easily create customizable, human-readable shortcuts to streamline your link management.
At a glance
- What is it?
- A self-hosted shortcut service that serves readable paths, supports private and public sharing, collections and tags, and resolves shortcuts from the browser search bar. The build is a pnpm frontend compiled into a static Go binary with a single data directory.
- Who is it for?
- Use Slash if you want readable internal links under your own domain, with private sharing for a team and public sharing for a collection, and you are happy to run one container. It is small enough to reason about: a single port, a single data directory, no external database in the documented path.
- Can I use it commercially?
- Yes, with strict conditions. AGPL-3.0 is a network copyleft licence: if people use a modified version over a network, for example as a hosted service, you must offer them its source code under the same licence.
- Is it still maintained?
- Yes. The repository last received commits 83 days ago.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- yourselfhosted/slash3,192 ★ · +14
| Repository | 2026-09-11 | 2026-11-15 |
|---|---|---|
| yourselfhosted/slash | 3.2k | 3.2k |
14 Stars in the last 65 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
One command, one port, one data directory
The deployment story is a single line, and everything else follows from it:
docker run -d --name slash -p 5231:5231 -v ~/.slash/:/var/opt/slash yourselfhosted/slash:latestThree things are decided by that command. The service listens on port 5231, which is also the port the image exposes and the one the compose file publishes. All state lives in a single host directory mounted at a path inside the container, and the container image creates and declares that path as a volume. The container is started detached with a name, so it is findable later.
That shape is the reason the project is small. There is no message queue, no cache and no separate database in the documented path, so a backup is a copy of one directory and a move is a copy of one directory.
The documentation splits the rest into three guides you would want in order: installing with Docker, getting started with shortcuts, and getting started with collections. A fourth page covers the browser extension.
The compose file pulls the published image, it does not build
The compose file is short enough to read in full, and one detail in it changes what you think you are running.
It declares a single service, gives it the published image with the latest tag, names the container, publishes the same port, mounts a named volume at the same in-container data path, and sets a restart policy of unless-stopped.
There is no build section. So `docker compose up` does not compile anything; it runs the image from the registry. That is the right default for most people and the wrong one if you have forked the code and expect your changes to appear.
There is a second difference worth noticing between the two paths. The documentation's run command bind-mounts a directory under your home directory, so the data is visible on the host immediately. The compose file uses a named volume instead, which Docker stores on its own volume directory. Both work; only the first one lets you point a backup tool at a path you already know.
The restart policy is the third quiet decision, and it is the one that matters overnight: the container comes back after a reboot without anyone logging in.
The frontend is built first, then embedded in the Go binary
The Dockerfile has three stages, and the middle one is where the design decision lives.
A Node stage builds the web application. It enables corepack, installs with a frozen lockfile so the build cannot silently drift from the committed dependency graph, and runs the package build script. A Go stage then compiles the server with cgo disabled, and before it compiles, it copies the built frontend distribution into a directory inside the server's route tree. So the static assets become part of the Go source tree at build time rather than being copied next to the binary at run time.
The final stage is a bare Alpine image with timezone data installed and the timezone environment variable pinned to UTC. It copies one file, the compiled binary, exposes the port, creates the data directory and declares it as a volume, sets a mode variable to production and a port variable, and uses the binary as its entry point. There is no command line, so the container's arguments go straight to the application.
One inconsistency is visible here. The Go image is pinned to the 1.23 line while the module file asks for a newer toolchain in its toolchain directive. If you build this yourself, the mismatch is the first thing to check.
Shortcuts, tags, collections and who can see them
The product is four ideas and a permission model.
A shortcut is a readable path under a short prefix, so a link becomes something like a single word after a slash instead of a query string. That is the whole premise: the point is that a human can remember it.
Tags categorise links, and they exist for the same reason shortcuts do, since a list of four hundred shortcuts needs a second axis.
Visibility is per link. A shortcut can be shared publicly or only with teammates, which means the same instance can hold an internal wiki and a public collection without two deployments.
Collections are the sharing unit: a set of shortcuts you hand to anyone, on any browser. That is the feature that makes the browser extension worth installing, because the recipient does not need an account to follow a collection.
Analytics exist on link traffic and sources, so you can tell which internal shortcuts are actually used, which is the measurement that turns a link list into a maintained one.
The extension resolves shortcuts from the search bar
A link shortener that only works when you are already on its own site saves you nothing, so the extension is the part that makes the rest worthwhile.
Its job is small and specific: you type a shortcut in the browser's search bar and it takes you to the corresponding URL. No new interface, no bookmark folder, just the omnibox you already use.
It is distributed the way extensions normally are, which is worth noting because it means the extension is not part of the self-hosted deployment. For Chromium based browsers, named in the documentation as Chrome, Edge and Arc among others, it comes from the Chrome Web Store. For Firefox it comes from the add-ons site.
So the split is clean: the server and its data are yours, and the client is a published extension that anyone can install. If you block store installs on your machines, you are back to typing the full URL, which is the state the project is trying to leave.
A small dependency surface and a legible module layout
The module layout tells you what the application is made of before you read a line of code.
There is a directory for the API definitions, one for extensions or plugins, one for storage, one for the server, one for the web frontend, and a bin directory whose main file is the single entry point the Dockerfile compiles. Beside them sit a docs directory and a scripts directory.
The dependency list is short for a service with a database, an API and a web UI. The web framework is one of the two most recognisable names in Go server development, the configuration library and the command line library are the other two standard picks, and there is a UUID library, the cryptography and networking packages from the standard extended set, and a testing library. Notably absent from the direct list is a database driver or an ORM, which matches the single data directory in the deployment.
The release configuration in the tree is also a GoReleaser file, which is the tool that produces the published image tags the compose file refers to.
Version 0.5.3 from February 2024, commits in August 2026
The version history is the last thing to look at, and it needs reading carefully.
The three newest releases are all from a two week window in early 2024: a minor release in late December, another in the third week of January, and the newest, 0.5.3, in the first week of February. Nothing has been tagged since.
The repository, though, is not idle. The last push in this snapshot is dated 2026-08-24, roughly two and a half years after that tag, and the project sits below version one.
That combination has a straightforward reading: the published image and the repository have drifted, which is exactly what you would expect from a small self-hosted project that fixes things as somebody reports them rather than shipping on a schedule. Nothing in the documentation claims otherwise, and the licence is the strong copyleft one, which is at least unambiguous about what you may do with it.
If you are pinning a deployment, pin the image tag rather than the latest label, because the latest label will move whenever somebody publishes.
Editorial conclusion
Use Slash if you want readable internal links under your own domain, with private sharing for a team and public sharing for a collection, and you are happy to run one container. It is small enough to reason about: a single port, a single data directory, no external database in the documented path. Skip it if you need a mature link management product with roles, previews and a large extension ecosystem, because the documented feature set is six lines long. Before you deploy, notice that the compose file pulls the published image rather than building your own, so if you want to build from source you need the build path, and remember the container pins its timezone rather than using yours. Check the release history before you depend on it, because the newest tag is from early 2024 while commits continue.
Frequently asked questions
How do I self-host Slash?
With one container command: run the published image detached, publish port 5231, and mount a host directory at /var/opt/slash. A compose file is provided as well, though it pulls the published image rather than building from source.
What port does Slash listen on?
Port 5231. It appears in the documented run command, in the compose file, as the exposed port in the image, and as the port environment variable set in the final build stage.
Does Slash need a separate database?
The documentation does not name one. State is kept in a single host directory mounted at /var/opt/slash, which the image creates and declares as a volume, and the direct dependencies include no database driver.
Can Slash links be private as well as public?
Yes. Short links can be shared publicly or only with teammates, and collections are the unit for sharing a set of shortcuts with anyone on any browser. Tags categorise links within the instance.
Is there a browser extension for Slash?
Yes, and it resolves shortcuts typed into the browser search bar. It is published in the Chrome Web Store for Chromium based browsers such as Chrome, Edge and Arc, and in the Firefox add-ons directory for Firefox.
What licence is Slash released under?
The repository is licensed under AGPL version 3, and the project describes itself as an open source self-hosted solution rather than a hosted service.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/yourselfhosted-slash)Star history chart for your README
[](https://hysenlabs.com/projects/yourselfhosted-slash)Was this useful?