Working with the AWS CDK in Python - AWS Cloud Development Kit (AWS CDK) v2

This is the AWS CDK v2 Developer Guide. The older CDK v1 entered maintenance on June 1, 2022 and will now receive only critical bug fixes and security patches. New features will be developed for CDK v2 exclusively. Support for CDK v1 will end entirely on June 1, 2023.

Working with the AWS CDK in Python

Python is a fully-supported client language for the AWS CDK and is considered stable. Working with the AWS CDK in Python uses familiar tools, including the standard Python implementation (CPython), virtual environments with virtualenv, and the Python package installer pip. The modules comprising the AWS Construct Library are distributed via The Python version of the AWS CDK even uses Python-style identifiers (for example, snake_case method names).

You can use any editor or IDE. Many AWS CDK developers use Visual Studio Code (or its open-source equivalent VSCodium), which has good support for Python via an official extension. The IDLE editor included with Python will suffice to get started. The Python modules for the AWS CDK do have type hints, which are useful for a linting tool or an IDE that supports type validation.


To work with the AWS CDK, you must have an AWS account and credentials and have installed Node.js and the AWS CDK Toolkit. See AWS CDK Prerequisites.

Python AWS CDK applications require Python 3.6 or later. If you don't already have it installed, download a compatible version for your operating system at If you run Linux, your system may have come with a compatible version, or you may install it using your distro's package manager (yum, apt, etc.). Mac users may be interested in Homebrew, a Linux-style package manager for macOS.


Third-party language deprecation: language version is only supported until its EOL (End Of Life) shared by the vendor or community and is subject to change with prior notice.

The Python package installer, pip, and virtual environment manager, virtualenv, are also required. Windows installations of compatible Python versions include these tools. On Linux, pip and virtualenv may be provided as separate packages in your package manager. Alternatively, you may install them with the following commands:

python -m ensurepip --upgrade python -m pip install --upgrade pip python -m pip install --upgrade virtualenv

If you encounter a permission error, run the above commands with the --user flag so that the modules are installed in your user directory, or use sudo to obtain the permissions to install the modules system-wide.


It is common for Linux distros to use the executable name python3 for Python 3.x, and have python refer to a Python 2.x installation. Some distros have an optional package you can install that makes the python command refer to Python 3. Failing that, you can adjust the command used to run your application by editing cdk.json in the project's main directory.


On Windows, you may want to invoke Python (and pip) using the py executable, the >Python launcher for Windows. Among other things, the launcher allows you to easily specify which installed version of Python you want to use.

If typing python at the command line results in a message about installing Python from the Windows Store, even after installing a Windows version of Python, open Windows' Manage App Execution Aliases settings panel and turn off the two App Installer entries for Python.

Creating a project

You create a new AWS CDK project by invoking cdk init in an empty directory.

mkdir my-project cd my-project cdk init app --language python

cdk init uses the name of the project folder to name various elements of the project, including classes, subfolders, and files. Hyphens in the folder name are converted to underscores. However, the name should otherwise follow the form of a Python identifier; for example, it should not start with a number or contain spaces.

To work with the new project, activate its virtual environment. This allows the project's dependencies to be installed locally in the project folder, instead of globally.

source .venv/bin/activate

You may recognize this as the Mac/Linux command to activate a virtual environment. The Python templates include a batch file, source.bat, that allows the same command to be used on Windows. The traditional Windows command, .venv\Scripts\activate.bat, works, too.

If you initialized your AWS CDK project using CDK Toolkit v1.70.0 or earlier, your virtual environment is in the .env directory instead of .venv.


Activate the project's virtual environment whenever you start working on it. Otherwise, you won't have access to the modules installed there, and modules you install will go in the Python global module directory (or will result in a permission error).

After activating your virtual environment for the first time, install the app's standard dependencies:

python -m pip install -r requirements.txt

Managing AWS Construct Library modules

Use the Python package installer, pip, to install and update AWS Construct Library modules for use by your apps, as well as other packages you need. pip also installs the dependencies for those modules automatically. If your system does not recognize pip as a standalone command, invoke pip as a Python module, like this:

python -m pip PIP-COMMAND

Most AWS CDK constructs are in aws-cdk-lib. Experimental modules are in separate modules named like aws-cdk.SERVICE-NAME.alpha. The service name includes an aws prefix. If you're unsure of a module's name, search for it at PyPI. For example, the command below installs the AWS CodeStar library.

python -m pip install

Some services' constructs are in more than one namespace. For example, besides, there are three additional Amazon RouteĀ 53 namespaces, named aws-route53-targets, aws-route53-patterns, and aws-route53resolver.


The Python edition of the CDK API Reference also shows the package names.

The names used for importing AWS Construct Library modules into your Python code look like the following.

import aws_cdk.aws_s3 as s3 import aws_cdk.aws_lambda as lambda_

We recommend the following practices when importing AWS CDK classes and AWS Construct Library modules in your applications. Following these guidelines will help make your code consistent with other AWS CDK applications as well as easier to understand.

  • Generally, import individual classes from top-level aws_cdk.

    from aws_cdk import App, Construct
  • If you need many classes from the aws_cdk, you may use a namespace alias of cdk instead of importing individual classes. Avoid doing both.

    import aws_cdk as cdk
  • Generally, import AWS Construct Libraries using short namespace aliases.

    import aws_cdk.aws_s3 as s3

After installing a module, update your project's requirements.txt file, which lists your project's dependencies. It is best to do this manually rather than using pip freeze. pip freeze captures the current versions of all modules installed in your Python virtual environment, which can be useful when bundling up a project to be run elsewhere.

Usually, though, your requirements.txt should list only top-level dependencies (modules that your app depends on directly) and not the dependencies of those libraries. This strategy makes updating your dependencies simpler.

You can edit requirements.txt to allow upgrades; simply replace the == preceding a version number with ~= to allow upgrades to a higher compatible version, or remove the version requirement entirely to specify the latest available version of the module.

With requirements.txt edited appropriately to allow upgrades, issue this command to upgrade your project's installed modules at any time:

pip install --upgrade -r requirements.txt

AWS CDK idioms in Python

Language conflicts

In Python, lambda is a language keyword, so you cannot use it as a name for the AWS Lambda construct library module or Lambda functions. The Python convention for such conflicts is to use a trailing underscore, as in lambda_, in the variable name.

By convention, the second argument to AWS CDK constructs is named id. When writing your own stacks and constructs, calling a parameter id "shadows" the Python built-in function id(), which returns an object's unique identifier. This function isn't used very often, but if you should happen to need it in your construct, rename the argument, for example construct_id.

Arguments and properties

All AWS Construct Library classes are instantiated using three arguments: the scope in which the construct is being defined (its parent in the construct tree), an id, and props, a bundle of key/value pairs that the construct uses to configure the resources it creates. Other classes and methods also use the "bundle of attributes" pattern for arguments.

scope and id should always be passed as positional arguments, not keyword arguments, because their names change if the construct accepts a property named scope or id.

In Python, props are expressed as keyword arguments. If an argument contains nested data structures, these are expressed using a class which takes its own keyword arguments at instantiation. The same pattern is applied to other method calls that take a structured argument.

For example, in a Amazon S3 bucket's add_lifecycle_rule method, the transitions property is a list of Transition instances.

bucket.add_lifecycle_rule( transitions=[ Transition( storage_class=StorageClass.GLACIER, transition_after=Duration.days(10) ) ] )

When extending a class or overriding a method, you may want to accept additional arguments for your own purposes that are not understood by the parent class. In this case you should accept the arguments you don't care about using the **kwargs idiom, and use keyword-only arguments to accept the arguments you're interested in. When calling the parent's constructor or the overridden method, pass only the arguments it is expecting (often just **kwargs). Passing arguments that the parent class or method doesn't expect results in an error.

class MyConstruct(Construct): def __init__(self, id, *, MyProperty=42, **kwargs): super().__init__(self, id, **kwargs) # ...

A future release of the AWS CDK could coincidentally add a new property with a name you used for your own property. This won't cause any technical issues for users of your construct or method (since your property isn't passed "up the chain," the parent class or overridden method will simply use a default value) but it may cause confusion. You can avoid this potential problem by naming your properties so they clearly belong to your construct. If there are many new properties, bundle them into an appropriately-named class and pass it as a single keyword argument.

Missing values

The AWS CDK uses None to represent missing or undefined values. When working with **kwargs, use the dictionary's get() method to provide a default value if a property is not provided. Avoid using kwargs[...], as this raises KeyError for missing values.

encrypted = kwargs.get("encrypted") # None if no property "encrypted" exists encrypted = kwargs.get("encrypted", False) # specify default of False if property is missing

Some AWS CDK methods (such as tryGetContext() to get a runtime context value) may return None, which you will need to check explicitly.

Using interfaces

Python doesn't have an interface feature as some other languages do, though it does have abstract base classes, which are similar. (If you're not familiar with interfaces, Wikipedia has a good introduction.) TypeScript, the language in which the AWS CDK is implemented, does provide interfaces, and constructs and other AWS CDK objects often require an object that adheres to a particular interface, rather than inheriting from a particular class. So the AWS CDK provides its own interface feature as part of the JSII layer.

To indicate that a class implements a particular interface, you can use the @jsii.implements decorator:

from aws_cdk import IAspect, IConstruct import jsii @jsii.implements(IAspect) class MyAspect(): def visit(self, node: IConstruct) -> None: print("Visited", node.node.path)

Type pitfalls

Python uses dynamic typing, where all variables may refer to a value of any type. Parameters and return values may be annotated with types, but these are "hints" and are not enforced. This means that in Python, it is easy to pass the incorrect type of value to a AWS CDK construct. Instead of getting a type error during build, as you would from a statically-typed language, you may instead get a runtime error when the JSII layer (which translates between Python and the AWS CDK's TypeScript core) is unable to deal with the unexpected type.

In our experience, the type errors Python programmers make tend to fall into these categories.

  • Passing a single value where a construct expects a container (Python list or dictionary) or vice versa.

  • Passing a value of a type associated with a layer 1 (CfnXxxxxx) construct to a L2 or L3 construct, or vice versa.

The AWS CDK Python modules do include type annotations, so you can use tools that support them to help with types. If you are not using an IDE that supports these, such as PyCharm, you might want to call the MyPy type validator as a step in your build process. There are also runtime type checkers that can improve error messages for type-related errors.

Synthesizing and deploying

The stacks defined in your AWS CDK app can be synthesized and deployed individually or together using the commands below. Generally, you should be in your project's main directory when you issue them.

  • cdk synth: Synthesizes a AWS CloudFormation template from one or more of the stacks in your AWS CDK app.

  • cdk deploy: Deploys the resources defined by one or more of the stacks in your AWS CDK app to AWS.

You can specify the names of multiple stacks to be synthesized or deployed in a single command. If your app defines only one stack, you do not need to specify it.

cdk synth # app defines single stack cdk deploy Happy Grumpy # app defines two or more stacks; two are deployed

You may also use the wildcards * (any number of characters) and ? (any single character) to identify stacks by pattern. When using wildcards, enclose the pattern in quotes. Otherwise, the shell may try to expand it to the names of files in the current directory before they are passed to the AWS CDK Toolkit.

cdk synth "Stack?" # Stack1, StackA, etc. cdk deploy "*Stack" # PipeStack, LambdaStack, etc.

You don't need to explicitly synthesize stacks before deploying them; cdk deploy performs this step for you to make sure your latest code gets deployed.

For full documentation of the cdk command, see AWS CDK Toolkit (cdk command).