git-open's manifest says 3.1.0, its newest tag says 2.1.0, and the issue flag has two short forms in two sections
Type `git open` to open the GitHub page or website for a repository in your browser.
At a glance
- What is it?
- git-open is a single Bash script that opens a repository's web page for ten hosting services, installed nine different ways and tested with a shell testing framework. Its own files disagree with each other in several places: the version in the manifest matches no tag, the changelog in the readme stops at the newest tag, the examples end mid-URL, a flag appears in the examples without appearing in the usage list, and the shell linter is defined but not wired into the test command.
- Who is it for?
- Use git-open if you want one shell verb for opening the page of whatever you are standing in, and if you already use Git from a terminal; the alternative projects listed in its own readme are worth checking first, because one of them is the tool GitHub maintains and another is the project this one was forked from. Three things to check before you install it.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 175 days ago.
- What is it written in?
- Mainly Shell, 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
- paulirish/git-open3,458 ★ · -2
| Repository | 2026-09-24 | 2026-11-15 |
|---|---|---|
| paulirish/git-open | 3.5k | 3.5k |
-2 Stars in the last 52 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
The manifest says 3.1.0, the newest tag says 2.1.0, and the changelog stops there
Three places in this repository record a version, and they do not agree.
The manifest declares a version three point one point zero. The newest tag is two point one point zero, published at the end of 2018. Before that comes a two point zero tag titled with a phrase about a big chill, and before that a one point one point one tag with no prefix at all.
So the manifest has run a full major version past the newest release, and one of the three tags in the visible history breaks the naming convention the other two follow.
Now look at the changelog. The readme has a changelog section with twelve entries, and the last one is the two point one point zero release. There is nothing about three point one point zero in it, and nothing about anything after December 2018.
There is also a separate changelog file at the repository root, in lower case, sitting beside a readme whose own changelog section is in title case. Two changelogs, one of them inside the readme and stopping at the newest tag.
Which means the manifest's version has never been released and the readme's changelog cannot tell you what changed since 2018, while the branch has been pushed to within the last few months.
One flag has two different short forms in two sections
The usage section gives every flag a long form and a short one, and the changelog gives a different short form for one of them.
The usage list shows an issue flag with the short form letter i. Its description is specific: if the branch is named like an issue followed by a number, it opens the corresponding issue on the repository's web page.
git open [remote-name] [branch-name]
# Open the page for this branch on the repo websiteThe changelog, near the bottom of the readme, records that the issue flag and another flag were added, and names the second one as letter h.
So the same flag is documented with one short form in the section people read and another in the section they scroll to.
There is a plausible innocent explanation: letter h was the original short form and letter i replaced it, with the changelog recording the addition rather than the change. That is exactly the case where a changelog entry and a usage list drift apart, and it is the case where a user typing the documented short form gets an error with no explanation.
The parenthetical note under the usage list compounds it slightly by saying the issue flag only works on three of the ten supported hosts, so the flag that has the ambiguous short form is also the host-restricted one.
The examples end mid-URL and document a flag the usage list omits
The examples section is six worked invocations, and the sixth one stops in the middle of its output.
The first five walk through the cases: no arguments, a named remote, a named remote and a named branch, the issue flag against a branch named after an issue, and the print flag. Each one is followed by the URL it produces, with placeholders for the parts that depend on your checkout.
The sixth passes a flag called suffix with the value for pull requests. And the comment beneath it gives the answer as two characters before the section ends.
Two things about that. The flag appears nowhere in the usage list above, so it is documented only by example. And the example's own output is cut off, which means the URL shape for that flag is not shown.
For a script whose entire job is constructing URLs, the one example that shows how to construct a URL for a pull request is the one you cannot read. Everything above it is complete.
Nine installation routes, and two of them run a Bash script on Windows
Count the documented ways to install this one file.
One is to put the script somewhere on your path, with the readme suggesting either adding its directory to the path variable or copying it into a system binary directory. One is a package manager install, with the manifest marked as preferring a global install. One is Windows through a shell profile: define a function that invokes the bash executable that ships with Git, and optionally give that function a short alias. One is Windows through a command prompt: put the script anywhere your path variable can reach.
Then four routes for one shell, each a different plugin manager, with the Antigen and the generator-manager routes being a single line each and the two longer ones being clone, edit a file, source it.
And one more for a different shell, the same three-step shape.
That is nine. What the two Windows routes have in common is that neither is native: both run the Bash script through Git's bundled interpreter, or require you to have it on your path.
Which is consistent with the project being one shell script with no compiled component, and it is why the manifest's engine requirement for a package manager is not about the tool at all.
A shell linter is defined and never called by the test command
The manifest defines three linters and a test command, and the relationship between them is not what you would expect.
The three linters are: one for the readme, one for the manual page, and one for whitespace and editor settings across the script, the markdown files and the YAML files. The aggregate lint command runs those three in order.
The test command runs the unit suite and then the aggregate lint.
And there is a fourth script, a shell linter for the shell script itself, which none of those chains call.
So the file that is the entire project is the one file the test command does not check. A change to the script that breaks a shell convention, or uses a quoting pattern a shell linter would flag, passes the whole pipeline.
That is one line of omission in a manifest that is otherwise careful: the readme gets a linter with its own configuration file, the manual page gets a linter with the same configuration file, and the script gets a linter nobody runs.
The package installs a second command the readme never mentions
The manifest declares two commands, and they are the same file.
The first maps the name of the script to the script. The second maps a different name, one that suggests opening a user's profile page rather than a repository page, to the same path.
So installing the package by the documented package-manager route gives you two commands, one of which appears nowhere in the readme.
The readme's alternatives section is where the omission becomes interesting, because it lists another project that can open user profile pages as well, and singles it out as the difference. So a second project is credited with a capability that this project's own package manifest is quietly shipping.
Which raises a question the readme does not answer: does the second command do something different, or is it an alias left in place from a version where profile opening worked and no longer does? There is no documentation either way, and no flag in the usage list suggests a profile mode.
The alternative list points at this project's own ancestor
The alternatives section lists four projects, and one of them is where this one came from.
The first is a tool maintained by the hosting company this project supports most, credited as the official project for that host, and it is recommended for complete opening support on that host.
The second is a different implementation of the same command name, credited with working only on that host but being able to open user profile pages, which is the capability the second installed command appears to promise.
The third is another implementation of the same command name, credited with working on a few providers, with one specific behavioural difference noted: it opens the default view for one hosting service rather than the source view.
The fourth is the original project this one was forked from, credited in a separate thanks section with the phrase that it did the initial hard work.
So the lineage is explicit and the credit is generous. What is interesting is the order: the ancestor is listed last, among the alternatives, rather than first, where someone arriving from the original project would look for it.
Ten host entries, three of them the same product under two names
The supported-hosts list has ten entries, and it carries its own history in parentheses.
There is the main hosting service and its separate gist host, so two entries for one company. There is the hosted form of one service and the self-hosted form, which is a useful distinction rather than a duplicate.
Then the second hosting service and its server product, with the former name given in parentheses as a rename the list has not caught up with.
Then two entries that the readme itself treats as one thing elsewhere: a hosted team services name and an on-premises server product from the same vendor. Both appear in the list of ten, and both appear together in the parenthetical note about which hosts the issue flag works with, as though they were one.
The last two are a managed cloud code service and a smaller host.
So a list of ten is closer to eight distinct services, and three of them share two vendors. That is not a defect; a user reading the list is better served by a longer list than a shorter accurate one. But the naming is inconsistent, and the rename in the parentheses is the kind of thing that ages badly.
Editorial conclusion
Use git-open if you want one shell verb for opening the page of whatever you are standing in, and if you already use Git from a terminal; the alternative projects listed in its own readme are worth checking first, because one of them is the tool GitHub maintains and another is the project this one was forked from. Three things to check before you install it. How you will get it, because there are nine documented routes across four shell plugin managers, a package manager, a path entry and two Windows workarounds, and the Windows ones both run a Bash script rather than a native binary. Which flags you need, since the usage list and the examples do not agree on the short form of one flag and one example documents a flag the usage list never mentions. And what version you are actually running, because the manifest, the newest tag and the changelog in the readme all give different answers, and the repository's last push is dated 2026-05-25 with the newest tag from 2018.
Frequently asked questions
What is git-open?
It is a single Bash script that opens the web page for the repository you are standing in, for GitHub, GitLab, Bitbucket and seven other hosting entries. With no arguments it opens the page for the current branch on the tracked remote, and you can pass a remote name and a branch name to override both. It is MIT licensed, with copyright shared between two people.
How do I open my Git?
For a repository page, run the script with no arguments. For the current commit there is a flag and a short form. For a pull request page there is a suffix flag, which appears in the examples rather than in the usage list, and the example that shows it ends mid-URL. To only print the URL without opening a browser there is a print flag and a short form.
How do I install git-open?
Nine documented routes: put the script on your path, install it globally through a package manager, two Windows workarounds that both run the script through the Bash interpreter bundled with Git, four plugin managers for one shell of which two are a single line each, and a three-step route for a different shell. The manifest marks the package install as preferring a global install.
Which hosting services does git-open support?
The list has ten entries: a main hosting service and its separate gist host, a hosted service and its self-hosted form, another hosting service and its server product with the former name given in parentheses, two entries for the same vendor's hosted and on-premises team products, a managed cloud code service, and one other host. The issue-opening flag is documented as working on only three of them.
Is git-open open source software?
Yes. It is MIT licensed, and the licence statement names two copyright holders rather than one. The manifest's author field names only the second of them and lists three contributors, so the copyright statement and the package metadata disagree about who wrote it.
How do I test git-open?
With a Bash testing framework, installed from a package manager or through a submodule that pulls in the assertion libraries. The manifest runs the suite through the framework's runner or through an equivalent script, and a separate watch command re-runs it on changes using a file-watcher. The test command then runs the markdown linters and the editor settings linter, but not the shell linter that is also defined in the manifest.
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/paulirish-git-open)Star history chart for your README
[](https://hysenlabs.com/projects/paulirish-git-open)Was this useful?