A better
npm publish
- Interactive UI
- Ensures you are publishing from your release branch (
main
andmaster
by default) - Ensures the working directory is clean and that there are no unpulled changes
- Reinstalls dependencies to ensure your project works with the latest dependency tree
- Ensures your Node.js and npm versions are supported by the project and its dependencies
- Runs the tests
- Bumps the version in package.json and npm-shrinkwrap.json (if present) and creates a git tag
- Prevents accidental publishing of pre-release versions under the
latest
dist-tag - Publishes the new version to npm, optionally under a dist-tag
- Rolls back the project to its previous state in case publishing fails
- Pushes commits and tags (newly & previously created) to GitHub/GitLab
- Supports two-factor authentication
- Enables two-factor authentication on new repositories
(does not apply to external registries) - Opens a prefilled GitHub Releases draft after publish
- Warns about the possibility of extraneous files being published
- See exactly what will be executed with preview mode, without pushing or publishing anything remotely
- Supports GitHub Packages
- Supports npm 9 , Yarn (Classic and Berry), pnpm 8 , and Bun
- Monorepos are not supported.
- Custom registries are not supported (but could be with your help).
- CI is not an ideal environment for
np
. It's meant to be used locally as an interactive tool.
- Node.js 18 or later
- npm 9 or later
- Git 2.11 or later
npm install --global np
$ np --help
Usage
$ np <version>
Version can be:
patch | minor | major | prepatch | preminor | premajor | prerelease | 1.2.3
Options
--any-branch Allow publishing from any branch
--branch Name of the release branch (default: main | master)
--no-cleanup Skips cleanup of node_modules
--no-tests Skips tests
--yolo Skips cleanup and testing
--no-publish Skips publishing
--preview Show tasks without actually executing them
--tag Publish under a given dist-tag
--contents Subdirectory to publish
--no-release-draft Skips opening a GitHub release draft
--release-draft-only Only opens a GitHub release draft for the latest published version
--test-script Name of npm run script to run tests before publishing (default: test)
--no-2fa Don't enable 2FA on new packages (not recommended)
--message Version bump commit message, '%s' will be replaced with version (default: '%s' with npm and 'v%s' with yarn)
--package-manager Use a specific package manager (default: 'packageManager' field in package.json)
Examples
$ np
$ np patch
$ np 1.0.2
$ np 1.0.2-beta.3 --tag=beta
$ np 1.0.2-beta.3 --tag=beta --contents=dist
Run np
without arguments to launch the interactive UI that guides you through publishing a new version.
np
can be configured both globally and locally. When using the global np
binary, you can configure any of the CLI flags in either a .np-config.js
(as CJS), .np-config.cjs
, .np-config.mjs
, or .np-config.json
file in the home directory. When using the local np
binary, for example, in a npm run
script, you can configure np
by setting the flags in either a top-level np
field in package.json
or in one of the aforementioned file types in the project directory. If it exists, the local installation will always take precedence. This ensures any local config matches the version of np
it was designed for.
Currently, these are the flags you can configure:
anyBranch
- Allow publishing from any branch (false
by default).branch
- Name of the release branch (main
ormaster
by default).cleanup
- Cleanupnode_modules
(true
by default).tests
- Runnpm test
(true
by default).yolo
- Skip cleanup and testing (false
by default).publish
- Publish (true
by default).preview
- Show tasks without actually executing them (false
by default).tag
- Publish under a given dist-tag (latest
by default).contents
- Subdirectory to publish (.
by default).releaseDraft
- Open a GitHub release draft after releasing (true
by default).testScript
- Name of npm run script to run tests before publishing (test
by default).2fa
- Enable 2FA on new packages (true
by default) (setting this tofalse
is not recommended).message
- The commit message used for the version bump. Any%s
in the string will be replaced with the new version. By default, npm uses%s
and Yarn usesv%s
.packageManager
- Set the package manager to be used. Defaults to the packageManager field in package.json, so only use if you can't update package.json for some reason.
For example, this configures np
to use unit-test
as a test script, and to use dist
as the subdirectory to publish:
package.json
{
"name": "superb-package",
"np": {
"testScript": "unit-test",
"contents": "dist"
}
}
.np-config.json
{
"testScript": "unit-test",
"contents": "dist"
}
.np-config.js
or .np-config.cjs
module.exports = {
testScript: 'unit-test',
contents: 'dist'
};
.np-config.mjs
export default {
testScript: 'unit-test',
contents: 'dist'
};
Note: The global config only applies when using the global np
binary, and is never inherited when using a local binary.
You can use any of the test/version/publish related npm lifecycle hooks in your package.json to add extra behavior.
For example, here we build the documentation before tagging the release:
{
"name": "my-awesome-package",
"scripts": {
"version": "./build-docs && git add docs"
}
}
You can also add np
to a custom script in package.json
. This can be useful if you want all maintainers of a package to release the same way (Not forgetting to push Git tags, for example). However, you can't use publish
as name of your script because it's an npm defined lifecycle hook.
{
"name": "my-awesome-package",
"scripts": {
"release": "np"
},
"devDependencies": {
"np": "*"
}
}
If you want to run a user-defined test script before publishing instead of the normal npm test
or yarn test
, you can use --test-script
flag or the testScript
config. This can be useful when your normal test script is running with a --watch
flag or in case you want to run some specific tests (maybe on the packaged files) before publishing.
For example, np --test-script=publish-test
would run the publish-test
script instead of the default test
.
{
"name": "my-awesome-package",
"scripts": {
"test": "ava --watch",
"publish-test": "ava"
},
"devDependencies": {
"np": "*"
}
}
Set the sign-git-tag
npm config to have the Git tag signed:
$ npm config set sign-git-tag true
Or set the version-sign-git-tag
Yarn config:
$ yarn config set version-sign-git-tag true
You can use np
for packages that aren't publicly published to npm (perhaps installed from a private git repo).
Set "private": true
in your package.json
and the publishing step will be skipped. All other steps
including versioning and pushing tags will still be completed.
To publish scoped packages to the public registry, you need to set the access level to public
. You can do that by adding the following to your package.json
:
"publishConfig": {
"access": "public"
}
If publishing a scoped package for the first time, np
will prompt you to ask if you want to publish it publicly.
Note: When publishing a scoped package, the first ever version you publish has to be done interactively using np
. If not, you cannot use np
to publish future versions of the package.
To publish a private Org-scoped package, you need to set the access level to restricted
. You can do that by adding the following to your package.json
:
"publishConfig": {
"access": "restricted"
}
Set the registry
option in package.json to the URL of your registry:
"publishConfig": {
"registry": "https://my-internal-registry.local"
}
If a package manager is not set in package.json, via configuration (packageManager
), or via the CLI (--package-manager
), np
will attempt to infer the best package manager to use by looking for lockfiles. But it's recommended to set the packageManager
field in your package.json to be consistent with other tools. See also the corepack docs.
If you use a Continuous Integration server to publish your tagged commits, use the --no-publish
flag to skip the publishing step of np
.
To publish to gh-pages
(or any other branch that serves your static assets), install branchsite
, an np
-like CLI tool aimed to complement np
, and create an npm "post" hook that runs after np
.
npm install --save-dev branchsite
"scripts": {
"deploy": "np",
"postdeploy": "bs"
}
For new packages, start the version
field in package.json at 0.0.0
and let np
bump it to 1.0.0
or 0.1.0
when publishing.
To release a minor/patch version for an old major version, create a branch from the major version's git tag and run np
:
$ git checkout -b fix-old-bug v1.0.0 # Where 1.0.0 is the previous major version
# Create some commits…
$ git push --set-upstream origin HEAD
$ np patch --any-branch --tag=v1
If you're using macOS Sierra 10.12.2 or later, your SSH key passphrase is no longer stored into the keychain by default. This may cause the prerequisite
step to run forever because it prompts for your passphrase in the background. To fix this, add the following lines to your ~/.ssh/config
and run a simple Git command like git fetch
.
Host *
AddKeysToAgent yes
UseKeychain yes
If you're running into other issues when using SSH, please consult GitHub's support article.
The ignore strategy, either maintained in the files
-property in package.json
or in .npmignore
, is meant to help reduce the package size. To avoid broken packages caused by essential files being accidentally ignored, np
prints out all the new and unpublished files added to Git. Test files and other common files that are never published are not considered. np
assumes either a standard directory layout or a customized layout represented in the directories
property in package.json
.
If you get an error like this…
❯ Prerequisite check
✔ Ping npm registry
✔ Check npm version
✔ Check yarn version
✖ Verify user is authenticated
npm ERR! code E403
npm ERR! 403 Forbidden - GET https://registry.yarnpkg.com/-/package/my-awesome-package/collaborators?format=cli - Forbidden
…please check whether the command npm access list collaborators my-awesome-package
succeeds. If it doesn't, Yarn has overwritten your registry URL. To fix this, add the correct registry URL to package.json
:
"publishConfig": {
"registry": "https://registry.npmjs.org"
}