Skip to main content

Log into Cognito, assume an IAM role, directly access JWT's from your Cognito session, and use boto3 all with Cognito credentials

Project description

Easy way to log into Cognito, assume IAM roles, access Cognito JWT, and use boto3 with Cognito credentials


If you have come here looking for cognito-assume-role then you should update your code to use cognitoinator. cognito-assume-role is no longer maintained

Currently supports USER_SRP_AUTH and USER_PASSWORD_AUTH using enhanced Cognito auth flow. Custom auth flows or administrative auth are not currently supported although I suppose you could monkey patch the needed code. It was written using only public methods exposed by the boto3/botocore API to help ensure that changes to boto3 wont break any (what would be otherwise) monkey patched code. This should also maintain the ability to use this library for your boto3 calls that are not using Cognito within the same script by simply passing a normal boto3 client credential argument.


This module can insert a botocore.credentials.CredentialProvider into the provider chain. Using this provider we can assume an IAM role through get_credentials_for_identity(). To do this we provide three functions:

  • client (wraps boto3.client())
  • resource (wraps boto3.resource())
  • Session (wraps boto3.session.Session())

All three of these functions accept all normal boto3 args and kwargs plus some that are specific to this module. We provide three ways of providing the initial credentials.


Env vars

These will take affect before any other credential provider, including the standard env provider that looks for AWS_SECRET_ACCESS_KEY and AWS_ACCESS_KEY_ID. If one or more of the following non-optional variables are found in environ then we will automatically go to env based credential mapping

  • COGNITO_METADATA (Deserialized and passed as ClientMetadata in boto3.client("cognito-idp").initiate_auth()) - Optional
  • AWS_ROLE_ARN - Optional


Credential file locations, if not specified, will be resolved in the order of

  • direct arguments to client
  • ~/.aws/cognito_credentials file

Config files take the following form:

metadata={"foo": "bar"}

All values except for region and metadata are required if using a profile. Using a profile is done by passing the kwarg "cognito_profile=profile name" to client, Session, or resource.

Direct configuration

  "username": "myusername",
  "password": "***********",
  "app_id": "1234567890",
  "user_pool_id": "abcdefg",
  "identity_pool_id": "us-east-1:1234567890",
  "region": "us-east-1",
  "metadata": {"foo": "bar"},
  "auth_type": "user_srp"

Same rules apply for required values as when using a profile. Direct configuration is done by passing the config dictionary to kwarg cognito_config when creating a client, resource, or Session. Note that cognito_profile and cognito_config are mutually exclusive. Trying to use both at once will raise an Assertion exception.

Precedence of CredentialProviders

The order of resolution for credential providers remains unchanged except for setting environment variables for Cognito will take affect before any AWS credential environment variables.

Precedence of arguments Any value that can be defined in either an environment variable, explicitly passed as a kwarg ( passed to client, resource, or Session) or can be part of a config or profile is resolved in the following order:

  • explicit arguments
  • specified by config or profile
  • environment variables

Auth types

The client, resource, and Session functions also accept an argument of auth_type. This can be user_srp (default) or user_password.

Assuming a role

Creating a client and assuming a role using env config

from cognitoinator import client

client = boto3.client("s3")

Creating a client that uses a config

from cognitoinator import client

client = boto3.client("s3", profile="my_profile")

Using resource with env vars and specifying auth_type and region

from cognitoinator import resource

resource = boto3.resource("s3", auth_type="user_password", region_name="us-east-2")

Creating a session that we can reuse for multiple clients

from cognitoinator import Session
session = Session(auth_type="user_srp", region_name="us-east-2")
s3 = session.client("s3")
dynamo = resource("dynamodb")
table = dynamo.Table("my_table")

Important notes on assuming roles

  • passing a role arn directly only works when your Identity Pool is configured for "Use default role" in (See Identity Pool -> Edit -> Authentication Providers -> Authenticated role selection). When using "rules" or "token" you cannot directly pass a role.
  • Assuming a role requires using enhanced_auth_flow. This is a requirement of AWS's assume-role-with-web-identity
  • Cognitoinator automatically will append the sub of the Cognito user to the role_session_name. This causes the IAM user accessing resources to be identified by their sub. This is very handy for creating resource based policies with grants to specific Cognito users based on their role_session_name.

Getting a JWT

Basic use

If you do not want to assume a role but would still like to access cognito id token directly, for instance to make Appsync calls, you can use the TokenFetcher class. It provides the following properties:

  • tokens (dict): A dictionary containing id_token, access_token, token_expires, and refresh_token
  • id_token
  • access_token
  • refresh_token
  • token_expires

Methods: - fetch(): Updates and returns self.tokens

All properties are available upon instantiation. The constructor accepts the same kwargs as Session(), along with option "server (bool)". Setting "server=True" will start a background process to keep tokens refreshed automatically, which means that your tokens will always be up to date.


from cognitoinator import TokenFetcher

cognito_credentials = TokenFetcher()


Getting tokens from cognitoinator.Session

If creating a Session directly the cognito id, refresh, and access tokens, as well as the expires time are available as properties on the Session object. Tokens are stored in memory by default, but passing a file name as "token_cache=/file/path.txt" into Session() will write cause the tokens to be written to the specified file as JSON. Passing a path to a file that does not exist will raise a FileNotFoundError. Passing a path to a file that is not writeable will raise OSError. Properties to access tokens:

  • Session().id_token
  • Session().access_token
  • Session().refresh_token
  • Session().token_expires
  • Session().cognito_tokens (All of the above in a dict)

Because of how boto3 generates clients there is no way to access the "parent" session. This means that to use this feature you will need to create a Session() object and then create your clients/resources off of that Session(). Example:

from cognitoinator import Session

session = Session()
s3 = session.s3()

# Outputs 2020-09-19T23:17:28CDT
from cognitoinator import TokenFetcher

s = TokenFetcher()
# Strings shortened for brevity

Results in:

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

cognitoinator-0.2.2.tar.gz (14.9 kB view hashes)

Uploaded source

Built Distribution

cognitoinator-0.2.2-py2.py3-none-any.whl (16.8 kB view hashes)

Uploaded py2 py3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page