Skip to content
Jump helps you navigate faster by learning your habits. ✌️
Go Makefile
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
assets Cleanup the git branches from the demo Oct 27, 2019
cli Fix misspell reports Jan 30, 2019
cmd Show the top score with jump top --score, keeping the current interface Dec 6, 2019
config Fix the mistaken ~/..jump home dir Jul 24, 2019
fuzzy Introduce full-path search support Oct 29, 2018
shell Always hint a single entries Apr 24, 2018
LICENSE.txt Update the license for 2019 Mar 13, 2019
Makefile Disable CGO by default, so we don't end up with a dynamic executable Nov 29, 2019
Vagrantfile Lock the FPM version as the latest yields broken debian packages May 28, 2019
go.sum Add go.mod file Sep 3, 2018


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_0.23.0_amd64.deb
Fedora wget && sudo rpm -i jump-0.23.0-1.x86_64.rpm
Go go get


Jump needs to be integrated with the shell. For bash and zsh, the the line below in 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 insenstive.

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. Somethimes we're looking for a directory the isn't the most scored one at the moment. Imagine the 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

You can’t perform that action at this time.