January 13, 2021

1867 words 9 mins read



A library that allows you to easily mock out tests based on AWS infrastructure.

repo name spulec/moto
repo link https://github.com/spulec/moto
homepage http://docs.getmoto.org/en/latest/
language Python
size (curr.) 15923 kB
stars (curr.) 4272
created 2013-02-18
license Apache License 2.0

Moto - Mock AWS Services

Join the chat at https://gitter.im/awsmoto/Lobby

Build Status Coverage Status Docs PyPI PyPI - Python Version PyPI - Downloads Code style: black


To install moto for a specific service:

$ pip install moto[ec2,s3]

This will install Moto, and the dependencies required for that specific service.
If you don’t care about the number of dependencies, or if you want to mock many AWS services:

$ pip install moto[all]

Not all services might be covered, in which case you might see a warning:
moto 1.3.16 does not provide the extra 'service'.
You can ignore the warning, or simply install moto as is:

$ pip install moto

In a nutshell

Moto is a library that allows your tests to easily mock out AWS Services.

Imagine you have the following python code that you want to test:

import boto3

class MyModel(object):
    def __init__(self, name, value):
        self.name = name
        self.value = value

    def save(self):
        s3 = boto3.client('s3', region_name='us-east-1')
        s3.put_object(Bucket='mybucket', Key=self.name, Body=self.value)

Take a minute to think how you would have tested that in the past.

Now see how you could test it with Moto:

import boto3
from moto import mock_s3
from mymodule import MyModel

def test_my_model_save():
    conn = boto3.resource('s3', region_name='us-east-1')
    # We need to create the bucket since this is all in Moto's 'virtual' AWS account

    model_instance = MyModel('steve', 'is awesome')

    body = conn.Object('mybucket', 'steve').get()['Body'].read().decode("utf-8")

    assert body == 'is awesome'

With the decorator wrapping the test, all the calls to s3 are automatically mocked out. The mock keeps the state of the buckets and keys.

It gets even better! Moto isn’t just for Python code and it isn’t just for S3. Look at the standalone server mode for more information about running Moto with other languages. Here’s the status of the other AWS services implemented:

Service Name Decorator Development Status Comment
ACM @mock_acm all endpoints done
API Gateway @mock_apigateway core endpoints done
Application Autoscaling @mock_applicationautoscaling basic endpoints done
Athena @mock_athena core endpoints done
Autoscaling @mock_autoscaling core endpoints done
Cloudformation @mock_cloudformation core endpoints done
Cloudwatch @mock_cloudwatch basic endpoints done
CloudwatchEvents @mock_events all endpoints done
Cognito Identity @mock_cognitoidentity basic endpoints done
Cognito Identity Provider @mock_cognitoidp basic endpoints done
Config @mock_config basic + core endpoints done
Data Pipeline @mock_datapipeline basic endpoints done
DynamoDB @mock_dynamodb core endpoints done API 20111205. Deprecated.
DynamoDB2 @mock_dynamodb2 all endpoints + partial indexes API 20120810 (Latest)
EC2 @mock_ec2 core endpoints done
- AMI core endpoints done
- EBS core endpoints done
- Instances all endpoints done
- Security Groups core endpoints done
- Tags all endpoints done
ECR @mock_ecr basic endpoints done
ECS @mock_ecs basic endpoints done
ELB @mock_elb core endpoints done
ELBv2 @mock_elbv2 all endpoints done
EMR @mock_emr core endpoints done
Forecast @mock_forecast some core endpoints done
Glacier @mock_glacier core endpoints done
Glue @mock_glue core endpoints done
IAM @mock_iam core endpoints done
IoT @mock_iot core endpoints done
IoT data @mock_iotdata core endpoints done
Kinesis @mock_kinesis core endpoints done
KMS @mock_kms basic endpoints done
Lambda @mock_lambda basic endpoints done, requires docker
Logs @mock_logs basic endpoints done
Organizations @mock_organizations some core endpoints done
Polly @mock_polly all endpoints done
RDS @mock_rds core endpoints done
RDS2 @mock_rds2 core endpoints done
Redshift @mock_redshift core endpoints done
Route53 @mock_route53 core endpoints done
S3 @mock_s3 core endpoints done
SecretsManager @mock_secretsmanager basic endpoints done
SES @mock_ses all endpoints done
SNS @mock_sns all endpoints done
SQS @mock_sqs core endpoints done
SSM @mock_ssm core endpoints done
Step Functions @mock_stepfunctions core endpoints done
STS @mock_sts core endpoints done
SWF @mock_swf basic endpoints done
X-Ray @mock_xray all endpoints done

For a full list of endpoint implementation coverage

Another Example

Imagine you have a function that you use to launch new ec2 instances:

import boto3

def add_servers(ami_id, count):
    client = boto3.client('ec2', region_name='us-west-1')
    client.run_instances(ImageId=ami_id, MinCount=count, MaxCount=count)

To test it:

from . import add_servers
from moto import mock_ec2

def test_add_servers():
    add_servers('ami-1234abcd', 2)

    client = boto3.client('ec2', region_name='us-west-1')
    instances = client.describe_instances()['Reservations'][0]['Instances']
    assert len(instances) == 2
    instance1 = instances[0]
    assert instance1['ImageId'] == 'ami-1234abcd'

Using moto 1.0.X with boto2

moto 1.0.X mock decorators are defined for boto3 and do not work with boto2. Use the @mock_AWSSVC_deprecated to work with boto2.

Using moto with boto2

from moto import mock_ec2_deprecated
import boto

def test_something_with_ec2():
    ec2_conn = boto.ec2.connect_to_region('us-east-1')

When using both boto2 and boto3, one can do this to avoid confusion:

from moto import mock_ec2_deprecated as mock_ec2_b2
from moto import mock_ec2


All of the services can be used as a decorator, context manager, or in a raw form.


def test_my_model_save():
    # Create Bucket so that test can run
    conn = boto3.resource('s3', region_name='us-east-1')
    model_instance = MyModel('steve', 'is awesome')
    body = conn.Object('mybucket', 'steve').get()['Body'].read().decode()

    assert body == 'is awesome'

Context Manager

def test_my_model_save():
    with mock_s3():
        conn = boto3.resource('s3', region_name='us-east-1')
        model_instance = MyModel('steve', 'is awesome')
        body = conn.Object('mybucket', 'steve').get()['Body'].read().decode()

        assert body == 'is awesome'

Raw use

def test_my_model_save():
    mock = mock_s3()

    conn = boto3.resource('s3', region_name='us-east-1')

    model_instance = MyModel('steve', 'is awesome')

    assert conn.Object('mybucket', 'steve').get()['Body'].read().decode() == 'is awesome'


IAM-like Access Control

Moto also has the ability to authenticate and authorize actions, just like it’s done by IAM in AWS. This functionality can be enabled by either setting the INITIAL_NO_AUTH_ACTION_COUNT environment variable or using the set_initial_no_auth_action_count decorator. Note that the current implementation is very basic, see this file for more information.


If this environment variable is set, moto will skip performing any authentication as many times as the variable’s value, and only starts authenticating requests afterwards. If it is not set, it defaults to infinity, thus moto will never perform any authentication at all.


This is a decorator that works similarly to the environment variable, but the settings are only valid in the function’s scope. When the function returns, everything is restored.

def test_describe_instances_allowed():
    policy_document = {
        "Version": "2012-10-17",
        "Statement": [
                "Effect": "Allow",
                "Action": "ec2:Describe*",
                "Resource": "*"
    access_key = ...
    # create access key for an IAM user/assumed role that has the policy above.
    # this part should call __exactly__ 4 AWS actions, so that authentication and authorization starts exactly after this

    client = boto3.client('ec2', region_name='us-east-1',

    # if the IAM principal whose access key is used, does not have the permission to describe instances, this will fail
    instances = client.describe_instances()['Reservations'][0]['Instances']
    assert len(instances) == 0

See the related test suite for more examples.

Experimental: AWS Config Querying

For details about the experimental AWS Config support please see the AWS Config readme here.

There are some important caveats to be aware of when using moto:

Failure to follow these guidelines could result in your tests mutating your REAL infrastructure!

How do I avoid tests from mutating my real infrastructure?

You need to ensure that the mocks are actually in place. Changes made to recent versions of botocore have altered some of the mock behavior. In short, you need to ensure that you always do the following:

  1. Ensure that your tests have dummy environment variables set up:

     export AWS_ACCESS_KEY_ID='testing'
     export AWS_SECRET_ACCESS_KEY='testing'
     export AWS_SECURITY_TOKEN='testing'
     export AWS_SESSION_TOKEN='testing'
  2. VERY IMPORTANT: ensure that you have your mocks set up BEFORE your boto3 client is established. This can typically happen if you import a module that has a boto3 client instantiated outside of a function. See the pesky imports section below on how to work around this.

Example on usage?

If you are a user of pytest, you can leverage pytest fixtures to help set up your mocks and other AWS resources that you would need.

Here is an example:

def aws_credentials():
    """Mocked AWS Credentials for moto."""
    os.environ['AWS_ACCESS_KEY_ID'] = 'testing'
    os.environ['AWS_SECRET_ACCESS_KEY'] = 'testing'
    os.environ['AWS_SECURITY_TOKEN'] = 'testing'
    os.environ['AWS_SESSION_TOKEN'] = 'testing'

def s3(aws_credentials):
    with mock_s3():
        yield boto3.client('s3', region_name='us-east-1')

def sts(aws_credentials):
    with mock_sts():
        yield boto3.client('sts', region_name='us-east-1')

def cloudwatch(aws_credentials):
    with mock_cloudwatch():
        yield boto3.client('cloudwatch', region_name='us-east-1')

... etc.

In the code sample above, all of the AWS/mocked fixtures take in a parameter of aws_credentials, which sets the proper fake environment variables. The fake environment variables are used so that botocore doesn’t try to locate real credentials on your system.

Next, once you need to do anything with the mocked AWS environment, do something like:

def test_create_bucket(s3):
    # s3 is a fixture defined above that yields a boto3 s3 client.
    # Feel free to instantiate another boto3 S3 client -- Keep note of the region though.

    result = s3.list_buckets()
    assert len(result['Buckets']) == 1
    assert result['Buckets'][0]['Name'] == 'somebucket'

What about those pesky imports?

Recall earlier, it was mentioned that mocks should be established BEFORE the clients are set up. One way to avoid import issues is to make use of local Python imports – i.e. import the module inside of the unit test you want to run vs. importing at the top of the file.


def test_something(s3):
   from some.package.that.does.something.with.s3 import some_func # <-- Local import for unit test
   # ^^ Importing here ensures that the mock has been established.      

   some_func()  # The mock has been established from the "s3" pytest fixture, so this function that uses
                # a package-level S3 client will properly use the mock and not reach out to AWS.

Other caveats

For Tox, Travis CI, and other build systems, you might need to also perform a touch ~/.aws/credentials command before running the tests. As long as that file is present (empty preferably) and the environment variables above are set, you should be good to go.

Stand-alone Server Mode

Moto also has a stand-alone server mode. This allows you to utilize the backend structure of Moto even if you don’t use Python.

It uses flask, which isn’t a default dependency. You can install the server ‘extra’ package with:

pip install "moto[server]"

You can then start it running a service:

$ moto_server ec2
 * Running on

You can also pass the port:

$ moto_server ec2 -p3000
 * Running on

If you want to be able to use the server externally you can pass an IP address to bind to as a hostname or allow any of your external interfaces with

$ moto_server ec2 -H
 * Running on

Please be aware this might allow other network users to access your server.

Then go to localhost to see a list of running instances (it will be empty since you haven’t added any yet).

If you want to use boto with this (using the simpler decorators above instead is strongly encouraged), the easiest way is to create a boto config file (~/.boto) with the following values:

is_secure = False
https_validate_certificates = False
proxy_port = 5000
proxy =

If you want to use boto3 with this, you can pass an endpoint_url to the resource



The standalone server has some caveats with some services. The following services require that you update your hosts file for your code to work properly:

  1. s3-control

For the above services, this is required because the hostname is in the form of AWS_ACCOUNT_ID.localhost. As a result, you need to add that entry to your host file for your tests to function properly.


Releases are done from travisci. Fairly closely following this: https://docs.travis-ci.com/user/deployment/pypi/

  • Commits to master branch do a dev deploy to pypi.
  • Commits to a tag do a real deploy to pypi.
comments powered by Disqus