Skip to content
Switch branches/tags


Build Status Go Report Card

Jump integrates with your shell and learns about your navigational habits by keeping track of the directories you visit. It gives you the most visited directory for the shortest search term you type.



Jump comes in packages for the following platforms.

Platform Command
macOS brew install jump
Ubuntu wget && sudo dpkg -i jump_0.40.0_amd64.deb
Fedora wget && sudo rpm -i jump-0.40.0-1.x86_64.rpm
Void xbps-install -S jump
Nix nix-env -iA nixpkgs.jump
Go go get


Jump needs to be integrated with the shell. For bash and zsh, the line below needs to be in ~/.bashrc, ~/bash_profile or ~/.zshrc:

eval "$(jump shell)"

For fish shell, put the line below needs to be in ~/.config/fish/

status --is-interactive; and source (jump shell fish | psub)

Once integrated, jump will automatically monitor directory changes and start building an internal database.

But j is not my favourite letter!

This is fine, you can bind jump to z, with this:

eval "$(jump shell --bind=z)"

And now, you can use jump like z dir and it would just work! This is only an example, you can bind it to anything. If you are one of those persons that likes to type a lot with their fingers, you can do:

eval "$(jump shell --bind=goto)"

Voila! goto dir becomes a thing. The possibilities are endless!


Once integrated, jump introduces the j helper. It accepts only search terms and as a design goal there are no arguments to j. Whatever you give it, it's treated as search term.

Jump uses fuzzy matching to find the desired directory to jump to. This means that your search terms are patterns that match the desired directory approximately rather than exactly. Typing 2 to 5 consecutive characters of the directory name is all that jump needs to find it.

Regular jump

The default search behavior of jump is to fuzzy match the directory name of a score. The match is case insensitive.

If you visit the directory /Users/genadi/Development/rails/web-console often, you can jump to it by:

$ j wc      # or...
$ j webc    # or...
$ j console # or...
$ j b-c     # or...

Using jump is all about saving key strokes. However, if you made the effort to type a directory base name exactly, jump will try to find the exact match, rather than fuzzy search.

$ j web-console
$ pwd

Deep jump

Given the following directories:


Typing j site matches only the base names of the directories. The base name of /Users/genadi/Development/society/website is website, the same as the other absolute path above. The jump above will land on the most scrored path, which is the society one, however what if we wanted to land on the chaos website?

$ j ch site
$ pwd

This instructs jump to look for a site match inside that is preceded by a ch match in the parent directory. The search is normalized only on the last two parts of the target paths. This will ensure a better match, because of the shorter path to fuzzy match on.

There are no depth limitations though and a jump to /Users/genadi/Development/society/website can look like:

$ j dev soc web
$ pwd

In fact, every space passed to j is converted to an OS separator. The last search term can be expressed as:

$ j dev/soc/web
$ pwd

Reverse jump

Bad jumps happen. Sometimes we're looking for a directory that doesn't have the best score at the moment. Let's work with the following following jump database:


Typing j web would lead to:

$ j web
$ pwd

If we didn't expect this result, instead of another search term, typing j without any arguments will instruct jump to go the second best match.

$ j
$ pwd

Case sensitive jump

To trigger a case-sensitive search, use a term that has a capital letter.

$ j Dev
$ pwd

The jump will resolve to /Users/genadi/Development even if there is /Users/genadi/Development/dev-tools that scores better.

Is it like autojump or z?

Yes, it is! You can import your datafile from autojump or z with:

$ jump import

This will try z first then autojump, so you can even combine all the entries from both tools.

The command is safe to run on pre-existing jump database, because if an entry exist in jump already, it won't be imported and it's score will remain unchanged. You can be explicit and choose to import autojump or z with:

$ jump import autojump
$ jump import z

If you want to know more about the difference between Jump, z, and autojump, check-out this Twitter conversation.

Thanks! 🙌

Thank you for stopping by and showing your interest in Jump!