4.1.1. Installation¶

The Python install manager can be installed from the Microsoft Store app or downloaded and installed from python.org/downloads. The two versions are identical.

To install through the Store, simply click “Install”. After it has completed, open a terminal and type python to get started.

To install the file downloaded from python.org, either double-click and select “Install”, or run Add-AppxPackage <path to MSIX> in Windows PowerShell.

After installation, the python, py, and pymanager commands should be available. If you have existing installations of Python, or you have modified your PATH variable, you may need to remove them or undo the modifications. See Troubleshooting for more help with fixing non-working commands.

When you first install a runtime, you will likely be prompted to add a directory to your PATH. This is optional, if you prefer to use the py command, but is offered for those who prefer the full range of aliases (such as python3.15.exe) to be available. The directory will be %LocalAppData%\Python\bin by default, but may be customized by an administrator. Click Start and search for “Edit environment variables for your account” for the system settings page to add the path.

Each Python runtime you install will have its own directory for scripts. These also need to be added to PATH if you want to use them.

The Python install manager will be automatically updated to new releases. This does not affect any installs of Python runtimes. Uninstalling the Python install manager does not uninstall any Python runtimes.

If you are not able to install an MSIX in your context, for example, you are using automated deployment software that does not support it, or are targeting Windows Server 2019, see Advanced installation below for more information.

4.1.2. Basic use¶

The recommended command for launching Python is python, which will either launch the version requested by the script being launched, an active virtual environment, or the default installed version, which will be the latest stable release unless configured otherwise. If no version is specifically requested and no runtimes are installed at all, the current latest release will be installed automatically.

For all scenarios involving multiple runtime versions, the recommended command is py. This may be used anywhere in place of python or the older py.exe launcher. By default, py matches the behaviour of python, but also allows command line options to select a specific version as well as subcommands to manage installations. These are detailed below.

Because the py command may already be taken by the previous version, there is also an unambiguous pymanager command. Scripted installs that are intending to use Python install manager should consider using pymanager, due to the lower chance of encountering a conflict with existing installs. The only difference between the two commands is when running without any arguments: py will launch your default interpreter, while pymanager will display help (pymanager exec ... provides equivalent behaviour to py ...).

Each of these commands also has a windowed version that avoids creating a console window. These are pyw, pythonw and pywmanager. A python3 command is also included that mimics the python command. It is intended to catch accidental uses of the typical POSIX command on Windows, but is not meant to be widely used or recommended.

To launch your default runtime, run python or py with the arguments you want to be passed to the runtime (such as script files or the module to launch):

$> py
...
$> python my-script.py
...
$> py -m this
...

The default runtime can be overridden with the PYTHON_MANAGER_DEFAULT environment variable, or a configuration file. See Configuration for information about configuration settings.

To launch a specific runtime, the py command accepts a -V:<TAG> option. This option must be specified before any others. The tag is part or all of the identifier for the runtime; for those from the CPython team, it looks like the version, potentially with the platform. For compatibility, the V: may be omitted in cases where the tag refers to an official release and starts with 3.

$> py -V:3.15 ...
$> py -V:3-arm64 ...

Runtimes from other distributors may require the company to be included as well. It should be separated from the tag by a slash (either / or \), and may be shortened to any prefix of its full value. Specifying the company is optional when it is PythonCore, and specifying the tag is optional (but not the slash) when you want the latest release from a specific company.

$> py -V:Distributor\1.0 ...
$> py -V:distrib/ ...

If no version is specified, but a script file is passed, the script will be inspected for a shebang line. This is a special format for the first line in a file that allows overriding the command. See Shebang lines for more information. When there is no shebang line, or it cannot be resolved, the script will be launched with the default runtime.

If you are running in an active virtual environment, have not requested a particular version, and there is no shebang line, the default runtime will be that virtual environment. In this scenario, the python command was likely already overridden and none of these checks occurred. However, this behaviour ensures that the py command can be used interchangeably.

When no runtimes are installed, any launch command will try to install the requested version and launch it. However, after any version is installed, only the py exec ... and pymanager exec ... commands will install if the requested version is absent. Other forms of commands will display an error and direct you to use py install first.

4.1.3. Command help¶

The py help command will display the full list of supported commands, along with their options. Any command may be passed the -? option to display its help, or its name passed to py help.

$> py help
$> py help install
$> py install /?

All commands support some common options, which will be shown by py help. These options must be specified after any subcommand. Specifying -v or --verbose will increase the amount of output shown, and -vv will increase it further for debugging purposes. Passing -q or --quiet will reduce output, and -qq will reduce it further.

The --config=<PATH> option allows specifying a configuration file to override multiple settings at once. See Configuration below for more information about these files.

4.1.4. Listing runtimes¶

$> py list [-f=|--format=<FMT>] [-1|--one] [--online|-s=|--source=<URL>] [<TAG>...]

The list of installed runtimes can be seen using py list. A filter may be added in the form of one or more tags (with or without company specifier), and each may include a <, <=, >= or > prefix to restrict to a range.

A range of formats are supported, and can be passed as the --format=<FMT> or -f <FMT> option. Formats include table (a user-friendly table view), csv (comma-separated table), json (a single JSON blob), jsonl (one JSON blob per result), exe (just the executable path), prefix (just the prefix path).

The --one or -1 option only displays a single result. If the default runtime is included, it will be the one. Otherwise, the “best” result is shown (“best” is deliberately vaguely defined, but will usually be the most recent version). The result shown by py list --one <TAG> will match the runtime that would be launched by py -V:<TAG>.

The --only-managed option excludes results that were not installed by the Python install manager. This is useful when determining which runtimes may be updated or uninstalled through the py command.

The --online option is short for passing --source=<URL> with the default source. Passing either of these options will search the online index for runtimes that can be installed. The result shown by py list --online --one <TAG> will match the runtime that would be installed by py install <TAG>.

$> py list --online 3.15

For compatibility with the old launcher, the --list, --list-paths, -0 and -0p commands (for example, py -0p) are retained. They do not allow additional options, and will produce legacy-formatted output.

4.1.5. Installing runtimes¶

$> py install [-s=|--source=<URL>] [-f|--force] [-u|--update] [--dry-run] [<TAG>...]

New runtime versions may be added using py install. One or more tags may be specified, and the special tag default may be used to select the default. Ranges are not supported for installation.

CPython publishes multiple runtime variants. For most users, PythonCore is recommended and provides the standard CPython runtime. Tags without any prefix are treated as PythonCore\<TAG>.

PythonTest includes the standard runtime together with the test suite and debugging symbols. It is intended for contributors or users who need these additional files for testing or debugging Python.

PythonEmbed provides a minimal embeddable distribution for applications that bundle Python rather than for interactive use. It would normally be used with the --target option described below.

To see the available runtime variants, run:

py list --online

To install a specific variant, include it in the tag passed to py install:

py install PythonTest\3.15

The --source=<URL> option allows overriding the online index that is used to obtain runtimes. This may be used with an offline index, as shown in Offline installs.

Passing --force will ignore any cached files and remove any existing install to replace it with the specified one.

Passing --update will replace existing installs if the new version is newer. Otherwise, they will be left. If no tags are provided with --update, all installs managed by the Python install manager will be updated if newer versions are available. Updates will remove any modifications made to the install, including globally installed packages, but will leave site-packages and Scripts directories as is. Virtual environments are also unaffected.

Passing --dry-run will generate output and logs, but will not modify any installs.

Passing --refresh will update all registrations for installed runtimes. This will recreate Start menu shortcuts, registry keys, and global aliases (such as python3.15.exe or for any installed scripts). These are automatically refreshed on installation of any runtime, but may need to be manually refreshed after installing packages.

In addition to the above options, the --target option will extract the runtime to the specified directory instead of doing a normal install. This is useful for embedding runtimes into larger applications. Unlike a normal install, py will not be aware of the extracted runtime, and no Start menu or other shortcuts will be created. To launch the runtime, directly execute the main executable (typically python.exe) in the target directory.

$> py install ... [-t=|--target=<PATH>] <TAG>

The py exec command will install the requested runtime if it is not already present. This is controlled by the automatic_install configuration (PYTHON_MANAGER_AUTOMATIC_INSTALL), and is enabled by default. If no runtimes are available at all, all launch commands will do an automatic install if the configuration setting allows. This is to ensure a good experience for new users, but should not generally be relied on rather than using the py exec command or explicit install commands.

4.1.6. Offline installs¶

To perform offline installs of Python, you will need to first create an offline index on a machine that has network access.

$> py install --download=<PATH> ... <TAG>...

The --download=<PATH> option will download the packages for the listed tags and create a directory containing them and an index.json file suitable for later installation. This entire directory can be moved to the offline machine and used to install one or more of the bundled runtimes:

$> py install --source="<PATH>\index.json" <TAG>...

The Python install manager can be installed by downloading its installer and moving it to another machine before installing.

Alternatively, the ZIP files in an offline index directory can simply be transferred to another machine and extracted. This will not register the install in any way, and so it must be launched by directly referencing the executables in the extracted directory, but it is sometimes a preferable approach in cases where installing the Python install manager is not possible or convenient.

In this way, Python runtimes can be installed and managed on a machine without access to the internet.

4.1.7. Uninstalling runtimes¶

$> py uninstall [-y|--yes] <TAG>...

Runtimes may be removed using the py uninstall command. One or more tags must be specified. Ranges are not supported here.

The --yes option bypasses the confirmation prompt before uninstalling.

Instead of passing tags individually, the --purge option may be specified. This will remove all runtimes managed by the Python install manager, including cleaning up the Start menu, registry, and any download caches. Runtimes that were not installed by the Python install manager will not be impacted, and neither will manually created configuration files.

$> py uninstall [-y|--yes] --purge

The Python install manager can be uninstalled through the Windows “Installed apps” settings page. This does not remove any runtimes, and they will still be usable, though the global python and py commands will be removed. Reinstalling the Python install manager will allow you to manage these runtimes again. To completely clean up all Python runtimes, run with --purge before uninstalling the Python install manager.

4.1.8. Configuration¶

Python install manager is configured with a hierarchy of configuration files, environment variables, command-line options, and registry settings. In general, configuration files have the ability to configure everything, including the location of other configuration files, while registry settings are administrator-only and will override configuration files. Command-line options override all other settings, but not every option is available.

This section will describe the defaults, but be aware that modified or overridden installs may resolve settings differently.

A global configuration file may be configured by an administrator and would be read first. The user configuration file is stored at %AppData%\Python\pymanager.json (note that this location is under Roaming, not Local) and is read next, overwriting any settings from earlier files. An additional configuration file may be specified as the PYTHON_MANAGER_CONFIG environment variable or the --config command line option (but not both). These locations may be modified by administrative customization options listed later.

The following settings are those that are considered likely to be modified in normal use. Later sections list those that are intended for administrative customization.

Standard configuration options

Dotted names should be nested inside JSON objects, for example, list.format would be specified as {"list": {"format": "table"}}.

4.1.9. Shebang lines¶

If the first line of a script file starts with #!, it is known as a “shebang” line. Linux and other Unix-like operating systems have native support for such lines and they are commonly used on such systems to indicate how a script should be executed. The python and py commands allow the same facilities to be used with Python scripts on Windows.

To allow shebang lines in Python scripts to be portable between Unix and Windows, a number of ‘virtual’ commands are supported to specify which interpreter to use. The supported virtual commands are:

• /usr/bin/env <ALIAS>

• /usr/bin/env -S <ALIAS>

• /usr/bin/<ALIAS>

• /usr/local/bin/<ALIAS>

• <ALIAS>

For example, if the first line of your script starts with

#! /usr/bin/python

The default Python or an active virtual environment will be located and used. As many Python scripts written to work on Unix will already have this line, you should find these scripts can be used by the launcher without modification. If you are writing a new script on Windows which you hope will be useful on Unix, you should use one of the shebang lines starting with /usr.

Any of the above virtual commands can have <ALIAS> replaced by an alias from an installed runtime. That is, any command generated in the global aliases directory (which you may have added to your PATH environment variable) can be used in a shebang, even if it is not on your PATH. This allows the use of shebangs like /usr/bin/python3.12 to select a particular runtime.

If no runtimes are installed, or if automatic installation is enabled, the requested runtime will be installed if necessary. See Configuration for information about configuration settings.

The /usr/bin/env form of shebang line will also search the PATH environment variable for unrecognized commands. This corresponds to the behaviour of the Unix env program, which performs the same search, but prefers launching known Python commands. A warning may be displayed when searching for arbitrary executables, and this search may be disabled by the shebang_can_run_anything configuration option.

Shebang lines that do not match any of the patterns are treated as Windows executable paths that are absolute or relative to the directory containing the script file. This is a convenience for Windows-only scripts, such as those generated by an installer, since the behavior is not compatible with Unix-style shells. These paths may be quoted, and may include multiple arguments, after which the path to the script and any additional arguments will be appended. This functionality may be disabled by the shebang_can_run_anything configuration option.

Since version 26.3 of the Python install manager, custom shebang templates may be added to your configuration file. Add the shebang_templates object with one member for each template (the string to match) and the command to use when the template is matched. Most commands should be py -V:<tag> (or pyw) to launch one of your installed runtimes. The py -3.<version> form is also allowed, as is a plain py to launch the default. No other arguments are supported.

{
    "shebang_templates": {
        "/usr/bin/python": "py",
        "/usr/bin/my_custom_python": "py -V:MyCustomPython/3"
    }
}

If the substitute command is not py or pyw, it will be written back into the shebang and regular handling continues. If launching arbitrary executables is permitted, then providing a full path will allow you to redirect from Python to any executable. The template should match either the entire line (ignoring leading and trailing whitespace), or up to the first space in the shebang line.

4.1.10. Advanced installation¶

For situations where an MSIX cannot be installed, such as some older administrative distribution platforms, there is an MSI available from the python.org downloads page. This MSI has no user interface, and can only perform per-machine installs to its default location in Program Files. It will attempt to modify the system PATH environment variable to include this install location, but be sure to validate this on your configuration.

Be aware that the MSI package does not bundle any runtimes, and so is not suitable for installs into offline environments without also creating an offline install index. See Offline installs and Administrative configuration for information on handling these scenarios.

Runtimes installed by the MSI are shared with those installed by the MSIX, and are all per-user only. The Python install manager does not support installing runtimes per-machine. To emulate a per-machine install, you can use py install --target=<shared location> as administrator and add your own system-wide modifications to PATH, the registry, or the Start menu.

When the MSIX is installed, but commands are not available in the PATH environment variable, they can be found under %LocalAppData%\Microsoft\WindowsApps\PythonSoftwareFoundation.PythonManager_3847v3x7pw1km or %LocalAppData%\Microsoft\WindowsApps\PythonSoftwareFoundation.PythonManager_qbz5n2kfra8p0, depending on whether it was installed from python.org or through the Windows Store. Attempting to run the executable directly from Program Files is not recommended.

To programmatically install the Python install manager, it is easiest to use WinGet, which is included with all supported versions of Windows:

$> winget install 9NQ7512CXL7T -e --accept-package-agreements --disable-interactivity

# Optionally run the configuration checker and accept all changes
$> py install --configure -y

To download the Python install manager and install on another machine, the following WinGet command will download the required files from the Store to your Downloads directory (add -d <location> to customize the output location). This also generates a YAML file that appears to be unnecessary, as the downloaded MSIX can be installed by launching or using the commands below.

$> winget download 9NQ7512CXL7T -e --skip-license --accept-package-agreements --accept-source-agreements

To programmatically install or uninstall an MSIX using only PowerShell, the Add-AppxPackage and Remove-AppxPackage PowerShell cmdlets are recommended:

$> Add-AppxPackage C:\Downloads\python-manager-25.0.msix
...
$> Get-AppxPackage PythonSoftwareFoundation.PythonManager | Remove-AppxPackage

The latest release can be downloaded and installed by Windows by passing the AppInstaller file to the Add-AppxPackage command. This installs using the MSIX on python.org, and is only recommended for cases where installing via the Store (interactively or using WinGet) is not possible.

$> Add-AppxPackage -AppInstallerFile https://www.python.org/ftp/python/pymanager/pymanager.appinstaller

Other tools and APIs may also be used to provision an MSIX package for all users on a machine, but Python does not consider this a supported scenario. We suggest looking into the PowerShell Add-AppxProvisionedPackage cmdlet, the native Windows PackageManager class, or the documentation and support for your deployment tool.

Regardless of the install method, users will still need to install their own copies of Python itself, as there is no way to trigger those installs without being a logged-in user. When using the MSIX, the latest version of Python will be available for all users to install without network access.

Note that the MSIX downloadable from the Store and from the Python website are subtly different and cannot be installed at the same time. Wherever possible, we suggest using the above WinGet commands to download the package from the Store to reduce the risk of setting up conflicting installs. There are no licensing restrictions on the Python install manager that would prevent using the Store package in this way.

4.1.11. Administrative configuration¶

There are a number of options that may be useful for administrators to override configuration of the Python install manager. These can be used to provide local caching, disable certain shortcut types, override bundled content. All of the above configuration options may be set, as well as those below.

Configuration options may be overridden in the registry by setting values under HKEY_LOCAL_MACHINE\Software\Policies\Python\PyManager, where the value name matches the configuration key and the value type is REG_SZ. Note that this key can itself be customized, but only by modifying the core config file distributed with the Python install manager. We recommend, however, that registry values are used only to set base_config to a JSON file containing the full set of overrides. Registry key overrides will replace any other configured setting, while base_config allows users to further modify settings they may need.

Note that most settings with environment variables support those variables because their default setting specifies the variable. If you override them, the environment variable will no longer work, unless you override it with another one. For example, the default value of confirm is literally %PYTHON_MANAGER_CONFIRM%, which will resolve the variable at load time. If you override the value to yes, then the environment variable will no longer be used. If you override the value to %CONFIRM%, then that environment variable will be used instead.

Configuration settings that are paths are interpreted as relative to the directory containing the configuration file that specified them.

Administrative configuration options

4.1.12. Installing free-threaded binaries¶

Pre-built distributions of the free-threaded build are available by installing tags with the t suffix.

$> py install 3.15t
$> py install 3.15t-arm64
$> py install 3.15t-32

This will install and register as normal. If you have no other runtimes installed, then python will launch this one. Otherwise, you will need to use py -V:3.15t or, if you have added the global aliases directory to your PATH environment variable, the python3.15t.exe commands.

4.1.13. Index signatures¶

Index files may be signed to detect tampering. A signature is a catalog file at the same URL as the index with .cat added to the filename. The catalog file should contain the hash of its matching index file, and should be signed with a valid Authenticode signature. This allows standard tooling (on Windows) to generate a signature, and any certificate may be used as long as the client operating system already trusts its certification authority (root CA).

Index signatures are only downloaded and checked when the local configuration’s source_settings section includes the index URL and requires_signature is true, or the index JSON contains requires_signature set to true. When the setting exists in local configuration, even when false, settings in the index are ignored.

As well as requiring a valid signature, the required_root_subject and required_publisher_subject settings can further restrict acceptable signatures based on the certificate Subject fields. Any attribute specified in the configuration must match the attribute in the certificate (additional attributes in the certificate are ignored). Typical attributes are CN= for the common name, O= for the organization, and C= for the publisher’s country.

Finally, the required_publisher_eku setting allows requiring that a specific Enhanced Key Usage (EKU) has been assigned to the publisher certificate. For example, the EKU 1.3.6.1.5.5.7.3.3 indicates that the certificate was intended for code signing (as opposed to server or client authentication). In combination with a specific root CA, this provides another mechanism to verify a legitimate signature.

This is an example source_settings section from a configuration file. In this case, the publisher of the feed is uniquely identified by the combination of the Microsoft Identity Verification root and the EKU assigned by that root. The signature for this case would be found at https://www.python.org/ftp/python/index-windows.json.cat.

{
  "source_settings": {
    "https://www.python.org/ftp/python/index-windows.json": {
      "requires_signature": true,
      "required_root_subject": "CN=Microsoft Identity Verification Root Certificate Authority 2020",
      "required_publisher_subject": "CN=Python Software Foundation",
      "required_publisher_eku": "1.3.6.1.4.1.311.97.608394634.79987812.305991749.578777327"
    }
  }
}

The same settings could be specified in the index.json file instead. In this case, the root and EKU are omitted, meaning that the signature must be valid and have a specific common name in the publisher’s certificate, but no other checks are used.

{
  "requires_signature": true,
  "required_publisher_subject": "CN=Python Software Foundation",
  "versions": [
    // ...
  ]
}

When settings from inside a feed are used, the user is notified and the settings are shown in the log file or verbose output. It is recommended to copy these settings into a local configuration file for feeds that will be used frequently, so that unauthorised modifications to the feed cannot disable verification.

It is not possible to override the location of the signature file in the feed or through a configuration file. Administrators can provide their own source_settings in a mandatory configuration file (see Administrative configuration).

If signature validation fails, you will be notified and prompted to continue. When interactive confirmation is not allowed (for example, because --yes was specified), it will always abort. To use a feed with invalid configuration in this scenario, you must provide a configuration file that disables signature checking for that feed.

"source_settings": {
  "https://www.example.com/feed-with-invalid-signature.json": {
    "requires_signature": false
  }
}

4.1.14. Proxy settings¶

By default, the Python install manager will use your system-wide proxy settings, including automatic authentication. For most users, this behaves the same as your browser and other applications.

To override the settings, use the NO_PROXY, HTTP_PROXY and HTTPS_PROXY environment variables. These should be set in the terminal session before using the Python install manager. Other applications may also use the variables, so use caution before setting them globally for your entire machine.

Setting NO_PROXY to any non-empty value will disable the use of any proxy. Use this to bypass your system’s default proxy setting and attempt to access the index server directly.

Setting HTTPS_PROXY to a string like example.com:8080 will connect to that proxy server for all HTTPS connections. For the default index, all connections will be using HTTPS, and so this is the usual setting to override.

The HTTP_PROXY variable may be needed if you are using a private index server that does not use encrypted connections, but does require the default proxy server to be overridden. It follows the same format as HTTPS_PROXY.

Credentials may be embedded in either setting, however, when HTTPS_PROXY has credentials embedded they will be used in preference to other credentials. Only one set of credentials will be used for both types of proxy.

4.1.15. Troubleshooting¶

If your Python install manager does not seem to be working correctly, work through these tests and fixes to see if it helps. If not, report an issue at our bug tracker, including any relevant log files (written to your %TEMP% directory by default).

Troubleshooting

4.2.1. Python application¶

An application written in Python does not necessarily require users to be aware of that fact. The embedded distribution may be used in this case to include a private version of Python in an install package. Depending on how transparent it should be (or conversely, how professional it should appear), there are two options.

Using a specialized executable as a launcher requires some coding, but provides the most transparent experience for users. With a customized launcher, there are no obvious indications that the program is running on Python: icons can be customized, company and version information can be specified, and file associations behave properly. In most cases, a custom launcher should simply be able to call Py_Main with a hard-coded command line.

The simpler approach is to provide a batch file or generated shortcut that directly calls the python.exe or pythonw.exe with the required command-line arguments. In this case, the application will appear to be Python and not its actual name, and users may have trouble distinguishing it from other running Python processes or file associations.

With the latter approach, packages should be installed as directories alongside the Python executable to ensure they are available on the path. With the specialized launcher, packages can be located in other locations as there is an opportunity to specify the search path before launching the application.

4.2.2. Embedding Python¶

Applications written in native code often require some form of scripting language, and the embedded Python distribution can be used for this purpose. In general, the majority of the application is in native code, and some part will either invoke python.exe or directly use python3.dll. For either case, extracting the embedded distribution to a subdirectory of the application installation is sufficient to provide a loadable Python interpreter.

As with the application use, packages can be installed to any location as there is an opportunity to specify search paths before initializing the interpreter. Otherwise, there are no fundamental differences between using the embedded distribution and a regular installation.

4.3.1. Free-threaded packages¶

Packages containing free-threaded binaries are named python-freethreaded for the 64-bit version, pythonx86-freethreaded for the 32-bit version, and pythonarm64-freethreaded for the ARM64 version. These packages contain both the python3.15t.exe and python.exe entry points, both of which run free-threaded.

4.9.1. PyWin32¶

The PyWin32 module by Mark Hammond is a collection of modules for advanced Windows-specific support. This includes utilities for:

• Component Object Model (COM)

• Win32 API calls

• Registry

• Event log

• Microsoft Foundation Classes (MFC) user interfaces

PythonWin is a sample MFC application shipped with PyWin32. It is an embeddable IDE with a built-in debugger.

4.9.2. cx_Freeze¶

cx_Freeze wraps Python scripts into executable Windows programs (*.exe files). When you have done this, you can distribute your application without requiring your users to install Python.

4.11.1. Installation steps¶

Four Python 3.15 installers are available for download - two each for the 32-bit and 64-bit versions of the interpreter. The web installer is a small initial download, and it will automatically download the required components as necessary. The offline installer includes the components necessary for a default installation and only requires an internet connection for optional features. See Installing without downloading for other ways to avoid downloading during installation.

After starting the installer, one of two options may be selected:

If you select “Install Now”:

• You will not need to be an administrator (unless a system update for the C Runtime Library is required or you install the Python install manager for all users)

• Python will be installed into your user directory

• The Python install manager will be installed according to the option at the bottom of the first page

• The standard library, test suite, launcher and pip will be installed

• If selected, the install directory will be added to your PATH

• Shortcuts will only be visible for the current user

Selecting “Customize installation” will allow you to select the features to install, the installation location and other options or post-install actions. To install debugging symbols or binaries, you will need to use this option.

To perform an all-users installation, you should select “Customize installation”. In this case:

• You may be required to provide administrative credentials or approval

• Python will be installed into the Program Files directory

• The Python install manager will be installed into the Windows directory

• Optional features may be selected during installation

• The standard library can be pre-compiled to bytecode

• If selected, the install directory will be added to the system PATH

• Shortcuts are available for all users

4.11.2. Removing the MAX_PATH limitation¶

Windows historically has limited path lengths to 260 characters. This meant that paths longer than this would not resolve and errors would result.

In the latest versions of Windows, this limitation can be expanded to approximately 32,000 characters. Your administrator will need to activate the “Enable Win32 long paths” group policy, or set LongPathsEnabled to 1 in the registry key HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem.

This allows the open() function, the os module and most other path functionality to accept and return paths longer than 260 characters.

After changing the above option, no further configuration is required.

4.11.3. Installing without UI¶

All of the options available in the installer UI can also be specified from the command line, allowing scripted installers to replicate an installation on many machines without user interaction. These options may also be set without suppressing the UI in order to change some of the defaults.

The following options (found by executing the installer with /?) can be passed into the installer:

All other options are passed as name=value, where the value is usually 0 to disable a feature, 1 to enable a feature, or a path. The full list of available options is shown below.

For example, to silently install a default, system-wide Python installation, you could use the following command (from an elevated command prompt):

python-3.9.0.exe /quiet InstallAllUsers=1 PrependPath=1 Include_test=0

To allow users to easily install a personal copy of Python without the test suite, you could provide a shortcut with the following command. This will display a simplified initial page and disallow customization:

python-3.9.0.exe InstallAllUsers=0 Include_launcher=0 Include_test=0
    SimpleInstall=1 SimpleInstallDescription="Just for me, no test suite."

(Note that omitting the launcher also omits file associations, and is only recommended for per-user installs when there is also a system-wide installation that included the launcher.)

The options listed above can also be provided in a file named unattend.xml alongside the executable. This file specifies a list of options and values. When a value is provided as an attribute, it will be converted to a number if possible. Values provided as element text are always left as strings. This example file sets the same options as the previous example:

<Options>
    <Option Name="InstallAllUsers" Value="no" />
    <Option Name="Include_launcher" Value="0" />
    <Option Name="Include_test" Value="no" />
    <Option Name="SimpleInstall" Value="yes" />
    <Option Name="SimpleInstallDescription">Just for me, no test suite</Option>
</Options>

4.11.4. Installing without downloading¶

As some features of Python are not included in the initial installer download, selecting those features may require an internet connection. To avoid this need, all possible components may be downloaded on-demand to create a complete layout that will no longer require an internet connection regardless of the selected features. Note that this download may be bigger than required, but where a large number of installations are going to be performed it is very useful to have a locally cached copy.

Execute the following command from Command Prompt to download all possible required files. Remember to substitute python-3.9.0.exe for the actual name of your installer, and to create layouts in their own directories to avoid collisions between files with the same name.

python-3.9.0.exe /layout [optional target directory]

You may also specify the /quiet option to hide the progress display.

4.11.5. Modifying an install¶

Once Python has been installed, you can add or remove features through the Programs and Features tool that is part of Windows. Select the Python entry and choose “Uninstall/Change” to open the installer in maintenance mode.

“Modify” allows you to add or remove features by modifying the checkboxes - unchanged checkboxes will not install or remove anything. Some options cannot be changed in this mode, such as the install directory; to modify these, you will need to remove and then reinstall Python completely.

“Repair” will verify all the files that should be installed using the current settings and replace any that have been removed or modified.

“Uninstall” will remove Python entirely, with the exception of the Python install manager, which has its own entry in Programs and Features.

4.11.6. Installing free-threaded binaries¶

To install pre-built binaries with free-threading enabled (see PEP 703), you should select “Customize installation”. The second page of options includes the “Download free-threaded binaries” checkbox.

Selecting this option will download and install additional binaries to the same location as the main Python install. The main executable is called python3.13t.exe, and other binaries either receive a t suffix or a full ABI suffix. Python source files and bundled third-party dependencies are shared with the main install.

The free-threaded version is registered as a regular Python install with the tag 3.13t (with a -32 or -arm64 suffix as normal for those platforms). This allows tools to discover it, and for the Python install manager to support py.exe -3.13t. Note that the launcher will interpret py.exe -3 (or a python3 shebang) as “the latest 3.x install”, which will prefer the free-threaded binaries over the regular ones, while py.exe -3.13 will not. If you use the short style of option, you may prefer to not install the free-threaded binaries at this time.

To specify the install option at the command line, use Include_freethreaded=1. See Installing without downloading for instructions on pre-emptively downloading the additional binaries for offline install. The options to include debug symbols and binaries also apply to the free-threaded builds.

Free-threaded binaries are also available on nuget.org.

4.12.1. Getting started¶

py

py -3.7

py -2

'py' is not recognized as an internal or external command,
operable program or batch file.

py --list

# Select any '3.*' tagged runtime
py -V:3

# Select any 'PythonCore' released runtime
py -V:PythonCore/

# Select PythonCore's latest Python 3 runtime
py -V:PythonCore/3

#! python
import sys
sys.stdout.write("hello from Python %s\n" % (sys.version,))

py hello.py

#! python3

4.12.2. Shebang lines¶

If the first line of a script file starts with #!, it is known as a “shebang” line. Linux and other Unix like operating systems have native support for such lines and they are commonly used on such systems to indicate how a script should be executed. This launcher allows the same facilities to be used with Python scripts on Windows and the examples above demonstrate their use.

To allow shebang lines in Python scripts to be portable between Unix and Windows, this launcher supports a number of ‘virtual’ commands to specify which interpreter to use. The supported virtual commands are:

• /usr/bin/env

• /usr/bin/python

• /usr/local/bin/python

• python

For example, if the first line of your script starts with

#! /usr/bin/python

The default Python or an active virtual environment will be located and used. As many Python scripts written to work on Unix will already have this line, you should find these scripts can be used by the launcher without modification. If you are writing a new script on Windows which you hope will be useful on Unix, you should use one of the shebang lines starting with /usr.

Any of the above virtual commands can be suffixed with an explicit version (either just the major version, or the major and minor version). Furthermore the 32-bit version can be requested by adding “-32” after the minor version. I.e. /usr/bin/python3.7-32 will request usage of the 32-bit Python 3.7. If a virtual environment is active, the version will be ignored and the environment will be used.

The /usr/bin/env form of shebang line has one further special property. Before looking for installed Python interpreters, this form will search the executable PATH for a Python executable matching the name provided as the first argument. This corresponds to the behaviour of the Unix env program, which performs a PATH search. If an executable matching the first argument after the env command cannot be found, but the argument starts with python, it will be handled as described for the other virtual commands. The environment variable PYLAUNCHER_NO_SEARCH_PATH may be set (to any value) to skip this search of PATH.

Shebang lines that do not match any of these patterns are looked up in the [commands] section of the launcher’s .INI file. This may be used to handle certain commands in a way that makes sense for your system. The name of the command must be a single argument (no spaces in the shebang executable), and the value substituted is the full path to the executable (additional arguments specified in the .INI will be quoted as part of the filename).

[commands]
/bin/xpython=C:\Program Files\XPython\python.exe

Any commands not found in the .INI file are treated as Windows executable paths that are absolute or relative to the directory containing the script file. This is a convenience for Windows-only scripts, such as those generated by an installer, since the behavior is not compatible with Unix-style shells. These paths may be quoted, and may include multiple arguments, after which the path to the script and any additional arguments will be appended.

4.12.3. Arguments in shebang lines¶

The shebang lines can also specify additional options to be passed to the Python interpreter. For example, if you have a shebang line:

#! /usr/bin/python -v

Then Python will be started with the -v option

4.12.4. Customization¶

[defaults]
python=3.7

[defaults]
python=3
python3=3.7

4.12.5. Diagnostics¶

If an environment variable PYLAUNCHER_DEBUG is set (to any value), the launcher will print diagnostic information to stderr (i.e. to the console). While this information manages to be simultaneously verbose and terse, it should allow you to see what versions of Python were located, why a particular version was chosen and the exact command-line used to execute the target Python. It is primarily intended for testing and debugging.

4.12.6. Dry run¶

If an environment variable PYLAUNCHER_DRYRUN is set (to any value), the launcher will output the command it would have run, but will not actually launch Python. This may be useful for tools that want to use the launcher to detect and then launch Python directly. Note that the command written to standard output is always encoded using UTF-8, and may not render correctly in the console.

4.12.7. Install on demand¶

If an environment variable PYLAUNCHER_ALLOW_INSTALL is set (to any value), and the requested Python version is not installed but is available on the Microsoft Store, the launcher will attempt to install it. This may require user interaction to complete, and you may need to run the command again.

An additional PYLAUNCHER_ALWAYS_INSTALL variable causes the launcher to always try to install Python, even if it is detected. This is mainly intended for testing (and should be used with PYLAUNCHER_DRYRUN).

4.12.8. Return codes¶

The following exit codes may be returned by the Python launcher. Unfortunately, there is no way to distinguish these from the exit code of Python itself.

The names of codes are as used in the sources, and are only for reference. There is no way to access or resolve them apart from reading this page. Entries are listed in alphabetical order of names.