Important: This documentation covers modern versions of Yarn.
For 1.x docs, see
yarn addyarn binyarn cache cleanyarn config getyarn config setyarn config unsetyarn configyarn constraints queryyarn constraints sourceyarn constraintsyarn dedupeyarn dlxyarn execyarn explain peer-requirementsyarn explainyarn infoyarn inityarn installyarn linkyarn nodeyarn npm audityarn npm infoyarn npm loginyarn npm logoutyarn npm publishyarn npm tag addyarn npm tag listyarn npm tag removeyarn npm whoamiyarn packyarn patch-commityarn patchyarn plugin import from sourcesyarn plugin importyarn plugin listyarn plugin removeyarn plugin runtimeyarn rebuildyarn removeyarn runyarn searchyarn set resolutionyarn set version from sourcesyarn set versionyarn stageyarn unlinkyarn unplugyarn upyarn upgrade-interactiveyarn version applyyarn version checkyarn versionyarn whyyarn workspaceyarn workspaces focusyarn workspaces foreachyarn workspaces list

yarn install

Install the project dependencies.


$> yarn install


Install the project :

yarn install

Validate a project when using Zero-Installs :

yarn install --immutable --immutable-cache

Validate a project when using Zero-Installs (slightly safer if you accept external PRs) :

yarn install --immutable --immutable-cache --check-cache




Format the output as an NDJSON stream


Abort with an error exit code if the lockfile was to be modified


Abort with an error exit code if the cache folder was to be modified


Always refetch the packages and ensure that their checksums are consistent


Verbosely print the output of the build steps of dependencies

--mode #0

Change what artifacts installs generate


This command sets up your project if needed. The installation is split into four different steps that each have their own characteristics:

  • Resolution: First the package manager will resolve your dependencies. The exact way a dependency version is privileged over another isn't standardized outside of the regular semver guarantees. If a package doesn't resolve to what you would expect, check that all dependencies are correctly declared (also check our website for more information: ).

  • Fetch: Then we download all the dependencies if needed, and make sure that they're all stored within our cache (check the value of cacheFolder in yarn config to see where the cache files are stored).

  • Link: Then we send the dependency tree information to internal plugins tasked with writing them on the disk in some form (for example by generating the .pnp.cjs file you might know).

  • Build: Once the dependency tree has been written on the disk, the package manager will now be free to run the build scripts for all packages that might need it, in a topological order compatible with the way they depend on one another. See for detail.

Note that running this command is not part of the recommended workflow. Yarn supports zero-installs, which means that as long as you store your cache and your .pnp.cjs file inside your repository, everything will work without requiring any install right after cloning your repository or switching branches.

If the --immutable option is set (defaults to true on CI), Yarn will abort with an error exit code if the lockfile was to be modified (other paths can be added using the immutablePatterns configuration setting). For backward compatibility we offer an alias under the name of --frozen-lockfile, but it will be removed in a later release.

If the --immutable-cache option is set, Yarn will abort with an error exit code if the cache folder was to be modified (either because files would be added, or because they'd be removed).

If the --check-cache option is set, Yarn will always refetch the packages and will ensure that their checksum matches what's 1/ described in the lockfile 2/ inside the existing cache files (if present). This is recommended as part of your CI workflow if you're both following the Zero-Installs model and accepting PRs from third-parties, as they'd otherwise have the ability to alter the checked-in packages before submitting them.

If the --inline-builds option is set, Yarn will verbosely print the output of the build steps of your dependencies (instead of writing them into individual files). This is likely useful mostly for debug purposes only when using Docker-like environments.

If the --mode=<mode> option is set, Yarn will change which artifacts are generated. The modes currently supported are:

  • skip-build will not run the build scripts at all. Note that this is different from setting enableScripts to false because the later will disable build scripts, and thus affect the content of the artifacts generated on disk, whereas the former will just disable the build step - but not the scripts themselves, which just won't run.

  • update-lockfile will skip the link step altogether, and only fetch packages that are missing from the lockfile (or that have no associated checksums). This mode is typically used by tools like Renovate or Dependabot to keep a lockfile up-to-date without incurring the full install cost.