Updated: 2025-06-12
(If you just want to know my solution, skip to Reading .desktop Files.)
Linux is a very powerful operating system that gets more user-friendly by the year, but it still has glaring holes. Today, I discovered there is no reliable method to finding out if a browser is installed.
Searching online reveals that most people only care about knowing if specific browsers are installed, or are content with the incomplete option of only checking for the most popular browsers (currently Chrome, Firefox, and Opera). The which command works fine combined with a resource like LinuxConfig.org’s List of Browsers available on Linux, but now your script requires periodic maintenance to add more browsers to the list1.
The next most common answer online is to see if xdg-open exists, based on the assumption that a browser must be installed if xdg-open is installed – this is simply false. You also can’t query xdg-open to see if it can handle URLs without opening a URL. If you use this method, you are effectively attacking the user with a pop-up2.
I’ve also found documentation for update-alternatives, but it requires knowing exactly what name it uses for browser entries. I was told to use x-www-browser, gnome-www-browser, and just www-browser. None of these work, and the command doesn’t allow you to list all of its entries because the --all option forces you into an interactive session to configure every single option it has. I refuse to go through that mess just to find how browsers are defined.
This should be simple, shouldn’t it?
Worse, I’ve also found multiple users stating that this fails to adhere to system defaults on Ubuntu, despite it being Ubuntu’s official default method.. and it doesn’t even WORK.
There’s also Debian Wiki’s DefaultWebBrowser which is somehow too dense and too sparse, and in light of what I’ve already tried – nonsensical. Suggesting the use of a $BROWSER environment variable strikes me as particularly insane6.
The best option I can find right now is to manually scan every menu entry for programs categorized with WebBrowser3. This still excludes some options, is convoluted, and has the possibility of failing due to mistakes in menu entries. But it’s the closest to a future-facing solution I could find.
(While asking an unrelated question on the Linux Mint Forums, I did get more information about browser defaults, which you may find useful.)
Reading .desktop Files
Update: xdg-settings get default-web-browser basically has already done what I was doing here5. If anything has a .desktop file for a browser, that should return whatever default has been selected, thus telling you a browser is installed.4
Menus in Linux use a standard format to describe programs/actions so they can be automatically organized. Even some CLI apps create these entries as well. These are organized by predefined Categories, including WebBrowser. While this may still exclude some niche browsers, the users of such browsers likely will know how to handle this incompatibility, and are very very rare.
The only issue with this solution is finding the files and having to parse them yourself. This is still a little tedious, but will last longer and detect new browsers automatically. Unfortunately, the only information I found on where these files should be is conflicting slightly. The Arch wiki says they are in 3 specific locations, but the actual spec excludes one of those and specifies an environment variable where the rest are located. So I check all of these locations, scan every file, and look for two things: 1) A Categories key with “WebBrowser” in it. 2) No Hidden key.
For my specific use case, since I only need to know a browser is installed, I can exit this search on the first hit, but obviously you can use the spec to actually find which browsers are installed, and even launch them or show their icons!
For reference’ sake, here’s what a .desktop file for LibreWolf could look like:
[Desktop Entry]
Type=Application
Name=LibreWolf
Comment=Privacy-focused web browser
Exec=/path/to/librewolf
Icon=/path/to/icon.png
Categories=Network;WebBrowser;
What the hell am I doing this for anyhow?
I’m writing a script to help me automatically select and install packages on a Linux system so that I don’t have to remember or constantly look up how to install things that aren’t part of the standard package management of my system, or the names of things I may have forgotten about because I use them rarely.
The fundamental design of this script includes the ability to run installs and configurations in any order, except where dependencies require a specific order. Part of it removes Firefox due to Mozilla’s recent privacy violations7, another part adds a replacement browser, but other parts of the script require a functioning browser to function. Initially, I thought I just needed a check to see if any browser is present, but later I realized I need to be able to count how many browsers are installed. While I’m glad I found out there is a simpler way to check for any browser, I’m still using a manual scan of .desktop files because I need to count them.
(As I write this, I realize that while my current solution addresses most of my needs, I should have each browser-requiring section of the script verify a browser is present, instead of requiring the user to briefly have two browsers installed at once. While I still think it’s better to not remove a browser until its replacement is installed just in case, there is a hole in my dependency management here – especially because the user can do their own things outside of my script, obviously.)
footnotes & updates
