requests_oauth2client is an OAuth 2.x client for Python, able to obtain, refresh and revoke tokens from any OAuth2.x/OIDC compliant Authorization Server. It sits upon and extends the famous requests HTTP client module.

It can act as an OAuth 2.0 / 2.1 client, to automatically get and renew Access Tokens, based on the Client Credentials, Authorization Code, Refresh token, Token Exchange, JWT Bearer, Device Authorization, Resource Owner Password or CIBA grants. Additional grant types are easy to add if needed.

It also supports OpenID Connect 1.0, PKCE, Client Assertions, Token Revocation and Introspection, Resource Indicators, JWT-secured Authorization Requests, Pushed Authorization Requests, Authorization Server Issuer Identification, Demonstrating Proof of Possession, as well as using custom params to any endpoint, and other important features that are often overlooked or needlessly complex in other client libraries.

And it also includes a wrapper around requests.Session that makes it super easy to use REST-style APIs, with or without OAuth 2.x.

Please note that despite the name, this library has no relationship with Google oauth2client library.

Full module documentation is available at https://guillp.github.io/requests_oauth2client/.

requests_oauth2client is available from PyPi, so installing it is as easy as:

pip install requests_oauth2client

Everything from requests_oauth2client is available from the root module, so you can import it like this:

from requests_oauth2client import *

Or you can import individual objects from this package as usual. Note that importing * automatically imports requests, so no need to import it yourself.

If you have already obtained an access token for the API you want to call, you can convert it to an instance of BearerToken. Instances of this class work as a requests compatible auth handler.

import requests
from requests_oauth2client import BearerToken

token = BearerToken("my_access_token")
resp = requests.get("https://my.protected.api/endpoint", auth=token)

This authentication handler will add a Authorization: Bearer <my_access_token> header in the request, with your access token value, properly formatted according to RFC6750.

OAuth2Client offers several methods that implement the communication to the various endpoints that are standardised by OAuth 2.0 and its extensions. These endpoints include the Token Endpoint, Revocation, Introspection, UserInfo, BackChannel Authentication and Device Authorization Endpoints.

You must provide the URLs for these endpoints if you intend to use them. Otherwise, only the Token Endpoint is mandatory to initialize an OAuth2Client.

To initialize an instance of OAuth2Client, you only need the Token Endpoint URI from your Authorization Server (AS), and the credentials for your application, typically a client_id and a client_secret, usually also provided by the AS:

from requests_oauth2client import OAuth2Client

oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    client_id="my_client_id",
    client_secret="my_client_secret",
)

The Token Endpoint is the only endpoint that is mandatory to obtain tokens. Credentials are used to authenticate the client everytime it sends a request to its Authorization Server. Usually, these are a static Client ID and Secret, which are the equivalent of a username and a password, but meant for an application instead of for a human user. The default authentication method used by OAuth2Client is Client Secret Post, but other standardized methods such as Client Secret Basic, Client Secret JWT or Private Key JWT are supported as well. See more about client authentication methods below.

Instead of providing each endpoint URL yourself, you may also use the AS metadata endpoint URI, or the document data itself, to initialize your OAuth 2.0 client with the appropriate endpoints.

OAuth2Client has dedicated methods to send requests to the Token Endpoint using different standardized grants. Since the Token Endpoint URL and Client Authentication Method are already declared for the client at initialization, the only required parameters for these methods are those that will be sent in the request to the Token Endpoint.

These methods directly return a BearerToken if the request is successful, or raise an exception if it fails. BearerToken contains all the data returned by the Token Endpoint, including the Access Token. It will also:

• Keep track of the Access Token expiration date (based on the expires_in hint as returned by the AS). This date is accessible with the expires_at attribute.
• Contain the Refresh Token, if returned by the AS, accessible with the refresh_token attribute.
• Contain the ID Token, if returned by the AS, accessible with the id_token attribute (typically available when using the Authorization Code flow).
• Keep track of other associated metadata as well, also accessible as attributes with the same name: token.custom_attr, or with subscription syntax token["my.custom.attr"].

You can create such a BearerToken yourself if needed:

from requests_oauth2client import BearerToken

bearer_token = BearerToken(access_token="an_access_token", expires_in=60)
print(bearer_token)
# {'access_token': 'an_access_token',
#  'expires_in': 55,
#  'token_type': 'Bearer'}
print(bearer_token.expires_at)
# datetime.datetime(2021, 8, 20, 9, 56, 59, 498793)
assert not bearer_token.is_expired()

print(bearer_token.expires_in)
# 40

Note that the expires_in indicator here is not static. It keeps track of the token lifetime, in seconds, and is calculated as the time flies. The actual static expiration date is accessible with the expires_at property. You can check if a token is expired with bearer_token.is_expired().

You can use a BearerToken instance anywhere you can use an access_token as string.

Using OAuth2Client directly is useful for testing or debugging OAuth2.x flows, but it may not be suitable for actual applications where tokens must be obtained, used during their lifetime, then obtained again or refreshed once they are expired. requests_oauth2client contains several requests compatible Auth Handlers (as subclasses of requests.auth.AuthBase), that will take care of obtaining tokens when required, then will cache those tokens until they are expired, and will obtain new ones (or refresh them, when possible), once the initial token is expired. Those are best used with a requests.Session, or an ApiClient, which is a wrapper around Session with a few enhancements as described below.

To send a request using the Client Credentials grant, use the .client_credentials() method, providing the necessary parameters as keyword arguments in the token request.

from requests_oauth2client import OAuth2Client

oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    client_id="client_id",
    client_secret="client_secret",
)

token = oauth2client.client_credentials(scope="myscope")
# or, if your AS uses resource indicator:
token = oauth2client.client_credentials(scope="myscope", resource="https://myapi.local")
# or, if your AS uses 'audience' as parameter to identify the requested API (Auth0 style):
token = oauth2client.client_credentials(audience="https://myapi.local")
# or, if your AS uses custom parameters:
token = oauth2client.client_credentials(scope="myscope", custom_param="custom_value")

Parameters such as scope, resource, or audience, as well as any other required parameters by the Authorization Server (AS), can be passed as keyword parameters. These parameters will be included in the token request sent to the AS. Please note that none of those parameters are mandatory at the client level, but some might be required by your AS to fulfill your request.

You can use the OAuth2ClientCredentialsAuth auth handler. It takes an OAuth2Client as parameter, and the additional kwargs to pass to the token endpoint:

import requests
from requests_oauth2client import OAuth2Client, OAuth2ClientCredentialsAuth

oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    client_id="client_id",
    client_secret="client_secret",
)

auth = OAuth2ClientCredentialsAuth(
    oauth2client, scope="myscope", resource="https://myapi.local"
)

# use it like this:
requests.get("https://myapi.local/resource", auth=auth)

# or like this:
session = requests.Session()
session.auth = auth

resp = session.get("https://myapi.local/resource")

Once again, extra parameters such as scope, resource or audience are allowed if required.

When you send your first request, OAuth2ClientCredentialsAuth will automatically retrieve an access token from the AS using the Client Credentials grant, then will include it in the request. Next requests will use the same token, as long as it is valid. A new token will be automatically retrieved once the previous one is expired.

You can configure a leeway, which is a period of time before the actual expiration, in seconds, when a new token will be obtained. This may help getting continuous access to the API when the client and API clocks are slightly out of sync. Use the parameter leeway to OAuth2ClientCredentialsAuth:

from requests_oauth2client import OAuth2ClientCredentialsAuth

auth = OAuth2ClientCredentialsAuth(
    oauth2client,
    scope="myscope",
    leeway=30,
)

Obtaining tokens using the Authorization code grant is made in 3 steps:

1. your application must open a specific url called the Authentication Request in a browser.
2. your application must obtain and validate the Authorization Response, which is a redirection back to your application that contains an Authorization Code as parameter. This redirect back (often called "callback") is initiated by the Authorization Server after any necessary interaction with the user is complete (Registration, Login, Profile completion, Multi-Factor Authentication, Authorization, Consent, etc.)
3. your application must then exchange this Authorization Code for an Access Token, with a request to the Token Endpoint.

Using an OAuth2Client will help you with all those steps, as described below.

To be able to use the Authorization Code grant, you need 2 (optionally 3) URIs:

• the URL for Authorization Endpoint, which is the url where you must send your Authorization Requests
• the Redirect URI, which is the url pointing to your application, where the Authorization Server will reply with Authorization Response
• optionally, the issuer identifier, if your AS uses Issuer Identification.

You can declare those URIs when initializing your OAuth2Client instance, or you can use the AS discovery endpoint to initialize those URLs automatically. Then you can generate valid Authorization Requests by calling the method .authorization_request(), with the request specific parameters, such as scope, state, nonce as parameter:

from requests_oauth2client import OAuth2Client

client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endoint",
    authorization_endpoint="https://url.to.the/authorization_endpoint",
    redirect_uri="https://url.to.my.application/redirect_uri",
    client_id="client_id",
    client_secret="client_secret",
)

az_request = client.authorization_request(scope="openid email profile")

print(az_request)
# this will look like this, with line feeds for display purposes only:
# https://url.to.the.as/authorization_endpoint
# ?client_id=client_id
# &redirect_uri=https%3A%2F%2Furl.to.my.application%2Fredirect_uri
# &response_type=code
# &scope=openid+email+profile
# &state=FBx9mWeLwoKGgG76vhi6v61-4mgxmgZhtWIa7aTffdY
# &nonce=iHZJokhkGOAojff1tdknRyz9mPZyy5vq9JDlVaUHyqk
# &code_challenge=TG7qgdyKnwUPuoQ6NNJRlLMoHbeVmJlB8g0VOcfQEkc
# &code_challenge_method=S256

# you can send the user to that url with:
import webbrowser

webbrowser.open(az_request.uri)

Note that the state, nonce and code_challenge parameters are generated with secure random values by default. Should you wish to use your own values, you can pass them as parameters to OAuth2Client.authorization_request(). For PKCE, you need to pass your generated code_verifier, and the code_challenge will automatically be derived from it. If you want to disable PKCE, you can pass code_challenge_method=None when initializing your OAuth2Client.

Once you have redirected the user browser to the Authorization Request URI, and after the user is successfully authenticated and authorized, plus any other extra interactive step is complete, the AS will respond with a redirection to your redirect_uri. That is the Authorization Response. It contains several parameters that must be retrieved by your client. The Authorization Code is one of those parameters, but you must also validate that the state matches your request; if using AS Issuer Identification, you must also validate that the issuer matches what is expected. You can do this with:

# using the `az_request` as defined above

response_uri = input(
    "Please enter the full url and/or params obtained on the redirect_uri: "
)
# say the callback url is https://url.to.my.application/redirect_uri?code=an_az_code&state=FBx9mWeLwoKGgG76vhi6v61-4mgxmgZhtWIa7aTffdY&issuer=https://url.to.the.as
az_response = az_request.validate_callback(response_uri)

This auth_response is an AuthorizationResponse instance and contains everything that is needed for your application to complete the authentication and get its tokens from the AS.

Once you have obtained the AS response, containing an authorization code, your application must exchange it for actual Token(s).

To exchange a code for Access and/or ID tokens, use the OAuth2Client.authorization_code() method. If you have obtained an AuthorizationResponse as described above, you can simply do:

token = oauth2client.authorization_code(az_response)

This will automatically include the code, redirect_uri and code_verifier parameters in the Token Request, as expected by the AS. You may include extra parameters if required, or you may pass your own parameters, without using an AuthorizationResponse instance, like this:

token = oauth2client.authorization_code(
    code=code,
    code_verifier=code_verifier,
    redirect_uri=redirect_uri,
    custom_param=custom_value,
)

The OAuth2AuthorizationCodeAuth handler takes an OAuth2Client and an authorization code as parameter, plus whatever additional keyword parameters are required by your Authorization Server:

from requests_oauth2client import OAuth2Client, ApiClient, OAuth2AuthorizationCodeAuth

oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    authorization_endpoint="https://url.to.the/authorization_endpoint",
    auth=("client_id", "client_secret"),
)

api_client = ApiClient(
    "https://your.protected.api/endpoint",
    auth=OAuth2AuthorizationCodeAuth(
        oauth2client,
        "my_authorization_code",
    ),
)

# any request using api_client will trigger exchanging the code for an access_token, which is then cached, and refreshed later if needed
resp = api_client.post(data={...})

OAuth2AuthorizationCodeAuth will take care of refreshing the token automatically once it is expired, using the refresh token, if available.

Authorization Requests generated by OAuth2Client.authorization_request() are instance of the class AuthorizationRequest. You can also use that class directly to generate your requests, but in that case you need to supply your Authorization Endpoint URI, your client_id, redirect_uri, etc. You can access every parameter from an AuthorizationRequest instance, as well as the generated code_verifier, as attributes of this instance. Once an Authorization Request URL is generated, it your application responsibility to redirect or otherwise send the user to that URL. You may use the webbrowser module from Python standard library to do so. Here is an example for generating Authorization Requests:

from requests_oauth2client import AuthorizationRequest

az_request = AuthorizationRequest(
    "https://url.to.the/authorization_endpoint",
    client_id="my_client_id",
    redirect_uri="http://localhost/callback",  # this redirect_uri is specific to your app
    scope="openid email profile",
    # extra parameters such as `resource` can be included as well if required by your AS
    resource="https://my.resource.local/api",
)
print(
    az_request
)  # this request will look like this, with line breaks for display purposes only
# https://url.to.the/authorization_endpoint
# ?client_id=my_client_id
# &redirect_uri=http%3A%2F%2Flocalhost%callback
# &response_type=code
# &state=kHWL4VwcbUbtPR4mtht6yMAGG_S-ZcBh5RxI_IGDmJc
# &nonce=mSGOS1M3LYU9ncTvvutoqUR4n1EtmaC_sQ3db4dyMAc
# &scope=openid+email+profile
# &code_challenge=W3n02f6xUKoDVbmhWEWz3h780b-Ci6ucnBS_d7nogmQ
# &code_challenge_method=S256
# &resource=https%3A%2F%2Fmy.resource.local%2Fapi

print(az_request.code_verifier)
# 'gYK-ZnQfoat2bghwed7oEz--wvn4D70ksJ5GuWO9sXXygZ7PMnUlSpBmMCcNRHxdgTS9m_roYwGxF6HQxIqZVwXmxRJUziFHUFxDrNuUIjCJCx6gBhPlpFbUXulB1fo2'

Helpers for the Device Authorization Grant are also included. To get device and user codes, read the response attributes (including Device Code, User Code, Verification URI, etc.), then pooling the Token Endpoint:

from requests_oauth2client import (
    OAuth2Client,
    DeviceAuthorizationPoolingJob,
    BearerToken,
)

client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    device_authorization_endpoint="https://url.to.the/device_authorization_endpoint",
    auth=("client_id", "client_secret"),
)

da_resp = client.authorize_device()

# `da_resp` contains the Device Code, User Code, Verification URI, and other info returned by the AS:
da_resp.device_code
da_resp.user_code
da_resp.verification_uri
da_resp.verification_uri_complete
da_resp.expires_at
da_resp.interval

# Send/show the Verification Uri and User Code to the user. They must use a browser to visit that URL, authenticate, and input the User Code.

# You can then request the Token endpoint to check if the user successfully authorized your device like this:
pool_job = DeviceAuthorizationPoolingJob(client, da_resp)

resp = None
while resp is None:
    resp = pool_job()

assert isinstance(resp, BearerToken)

DeviceAuthorizationPoolingJob will automatically obey the pooling period. Everytime you call pool_job(), it will wait the appropriate number of seconds as indicated by the AS, and will apply slow-down requests.

Use OAuth2DeviceCodeAuth as auth handler to exchange a device code for an access token:

from requests_oauth2client import ApiClient, OAuth2DeviceCodeAuth, OAuth2Client

client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    device_authorization_endpoint="https://url.to.the/device_authorization_endpoint",
    auth=("client_id", "client_secret"),
)

device_auth_resp = client.authorize_device()

# expose user_code and verification_uri or verification_uri_complete to the user
device_auth_resp.user_code
device_auth_resp.verification_uri
device_auth_resp.verification_uri_complete

# then try to send your request with an OAuth2DeviceCodeAuth handler
# this will pool the token endpoint until the user authorizes the device
api_client = ApiClient(
    "https://your.protected.api/endpoint",
    auth=OAuth2DeviceCodeAuth(client, device_auth_resp),
)

resp = api_client.post(
    data={...}
)  # the first call will hang until the user authorizes your app and the token endpoint returns a token.

To initiate a BackChannel Authentication against the dedicated endpoint, read the response attributes and pool the Token Endpoint until the end-user successfully authenticates:

from requests_oauth2client import (
    OAuth2Client,
    BearerToken,
    BackChannelAuthenticationPoolingJob,
)

client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    backchannel_authentication_endpoint="https://url.to.the/backchannel_authorization_endpoint",
    auth=("client_id", "client_secret"),
)

ba_resp = client.backchannel_authentication_request(
    scope="openid email profile",
    login_hint="[email protected]",
)

# `ba_resp` will contain the response attributes as returned by the AS, including an `auth_req_id`:
ba_resp.auth_req_id
ba_resp.expires_in  # decreases with time
ba_resp.expires_at  # a static `datetime` to keep track of the expiration date, based on the "expires_in" returned by the AS
ba_resp.interval  # the pooling interval indicated by the AS
ba_resp.custom  # if the AS respond with additional attributes, they are also accessible

pool_job = BackChannelAuthenticationPoolingJob(client, ba_resp)

resp = None
while resp is None:
    resp = pool_job()

assert isinstance(resp, BearerToken)

Hints by the AS to slow down pooling will automatically be obeyed.

To send a token exchange request, use the OAuth2Client.token_exchange() method:

from requests_oauth2client import OAuth2Client, ClientSecretJwt

client = OAuth2Client(
    "https://url.to.the/token_endpoint",
    auth=ClientSecretJwt("client_id", "client_secret"),
)
token = client.token_exchange(
    subject_token="your_token_value",
    subject_token_type="urn:ietf:params:oauth:token-type:access_token",
)

As with the other grant-type specific methods, you may specify additional keyword parameters, that will be passed to the token endpoint, including any standardised attribute like actor_token or actor_token_type, or any custom parameter. There are short names for token types, that will be automatically translated to standardised types:

token = client.token_exchange(
    subject_token="your_token_value",
    subject_token_type="access_token",  # will be automatically replaced by "urn:ietf:params:oauth:token-type:access_token"
    actor_token="your_actor_token",
    actor_token_type="id_token",  # will be automatically replaced by "urn:ietf:params:oauth:token-type:id_token"
)

Or to make it even easier, types can be guessed based on the supplied subject or actor token:

from requests_oauth2client import BearerToken, ClientSecretJwt, IdToken, OAuth2Client

client = OAuth2Client(
    "https://url.to.the/token_endpoint",
    auth=ClientSecretJwt("client_id", "client_secret"),
)

token = client.token_exchange(
    subject_token=BearerToken(
        "your_token_value"
    ),  # subject_token_type will be "urn:ietf:params:oauth:token-type:access_token"
    actor_token=IdToken(
        "your_actor_token"
    ),  # actor_token_type will be "urn:ietf:params:oauth:token-type:id_token"
)

requests_oauth2client supports several client authentication methods, as defined in multiple OAuth2.x standards. You select the appropriate method to use when initializing your OAuth2Client, with the auth parameter. Once initialized, a client will automatically use the configured authentication method every time it sends a requested to an endpoint that requires client authentication. You don't have anything else to do afterwards.

With client_secret_basic, client_id and client_secret are included in clear-text in the Authorization header when sending requests to the Token Endpoint. To use it, just pass a ClientSecretBasic(client_id, client_secret) as auth parameter:

from requests_oauth2client import OAuth2Client, ClientSecretBasic

client = OAuth2Client(
    "https://url.to.the/token_endpoint",
    auth=ClientSecretBasic("client_id", "client_secret"),
)

With client_secret_post, client_id and client_secret are included as part of the body form data. To use it, pass a ClientSecretPost(client_id, client_secret) as auth parameter. This is the default when you pass a tuple (client_id, client_secret) as auth when initializing an OAuth2Client:

from requests_oauth2client import OAuth2Client, ClientSecretPost

client = OAuth2Client(
    "https://url.to.the/token_endpoint",
    auth=ClientSecretPost("client_id", "client_secret"),
)
# or
client = OAuth2Client(
    "https://url.to.the/token_endpoint", auth=("client_id", "client_secret")
)
# or
oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    client_id="my_client_id",
    client_secret="my_client_secret",
)

With client_secret_jwt, the client generates an ephemeral JWT assertion including information about itself (client_id), the AS (url of the endpoint), and an expiration date a few seconds in the future. To use it, pass a ClientSecretJwt(client_id, client_secret) as auth parameter. Assertion generation is entirely automatic, you don't have anything to do:

from requests_oauth2client import OAuth2Client, ClientSecretJwt

client = OAuth2Client(
    "https://url.to.the/token_endpoint",
    auth=ClientSecretJwt("client_id", "client_secret"),
)

This method is more secure than the 2 previous, because only ephemeral credentials are transmitted, which limits the possibility for interception and replay of the Client Secret. But that Client Secret still needs to be shared between the AS and Client owner(s).

With private_key_jwt, client uses a JWT assertion that is just like the one for client_secret_jwt, but it is signed with an asymmetric key. To use it, you need a private signing key, in a dict that matches the JWK format, or as an instance of jwskate.Jwk. The matching public key must be registered for your client on AS side. Once you have that, using this auth method is simple with the PrivateKeyJwt(client_id, private_jwk) auth handler:

from requests_oauth2client import OAuth2Client, PrivateKeyJwt

private_jwk = {
    "kid": "mykid",
    "kty": "RSA",
    "e": "AQAB",
    "n": "...",
    "d": "...",
    "p": "...",
    "q": "...",
    "dp": "...",
    "dq": "...",
    "qi": "...",
}

client = OAuth2Client(
    "https://url.to.the/token_endpoint", auth=PrivateKeyJwt("client_id", private_jwk)
)
# or
client = OAuth2Client(
    "https://url.to.the/token_endpoint", auth=("client_id", private_jwk)
)
# or
client = OAuth2Client(
    "https://url.to.the/token_endpoint", client_id="client_id", private_jwk=private_jwk
)

This method can be considered more secure than those relying on a client secret, because only ephemeral credentials are sent over the wire, and it uses asymmetric cryptography: the signing key is generated by the client, and only the public key is known by the AS. Transmitting that public key between owner(s) of the client and of the AS is much easier than transmitting the Client Secret, which is a shared key that must be considered as confidential.

The latest Client Authentication Method, none, is for Public Clients which do not authenticate to the Token Endpoint. Those clients only include their client_id in body form data, without any authentication credentials. Use PublicApp(client_id):

from requests_oauth2client import OAuth2Client, PublicApp

client = OAuth2Client(
    "https://url.to.the/token_endpoint", auth=PublicApp("app_client_id")
)

The OAuth2Client class provides methods for sending revocation requests to a Revocation Endpoint. To use this feature, you need to provide the Revocation Endpoint URI when creating an instance of OAuth2Client. The available methods for revoking tokens are:

• revoke_token(): Revokes a token by providing the token value and an optional token_type_hint.
• revoke_access_token(): Revokes an access token by providing the token value.
• revoke_refresh_token(): Revokes a refresh token by providing the token value.

Here is an example of how to use these methods:

from requests_oauth2client import OAuth2Client, ClientSecretJwt

oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    revocation_endpoint="https://url.to.the/revocation_endpoint",
    auth=ClientSecretJwt("client_id", "client_secret"),
)

oauth2client.revoke_token("mytoken", token_type_hint="access_token")
oauth2client.revoke_access_token("mytoken")
oauth2client.revoke_refresh_token("mytoken")

These methods return a boolean value indicating whether the revocation request was successfully sent and no error was returned. If the Authorization Server returns a non-successful HTTP code without a standard error message, it will return False. If the Authorization Server returns a standard error, an exception will be raised.

The OAuth2Client class also supports sending requests to a Token Introspection Endpoint. To use this feature, you need to provide the Introspection Endpoint URI when creating an instance of OAuth2Client. The introspect_token() method is then available for introspecting tokens:

from requests_oauth2client import OAuth2Client, ClientSecretJwt

oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    introspection_endpoint="https://url.to.the/introspection_endpoint",
    auth=ClientSecretJwt("client_id", "client_secret"),
)

resp = oauth2client.introspect_token("mytoken", token_type_hint="access_token")

The introspect_token() method returns the data returned by the introspection endpoint, decoded if it is in JSON format.

The OAuth2Client class also supports sending requests to a UserInfo Endpoint. To use this feature, you need to provide the UserInfo Endpoint URI when creating an instance of OAuth2Client The userinfo() method is then available for retrieving user information:

from requests_oauth2client import OAuth2Client, ClientSecretJwt

oauth2client = OAuth2Client(
    token_endpoint="https://url.to.the/token_endpoint",
    userinfo_endpoint="https://url.to.the/userinfo_endpoint",
    auth=ClientSecretJwt("client_id", "client_secret"),
)

resp = oauth2client.userinfo("mytoken")

The userinfo() method returns the data returned by the userinfo endpoint, decoded if it is in JSON format.

You can initialize an OAuth2Client with the endpoint URIs mentioned in a standardised discovery document using the OAuth2Client.from_discovery_endpoint() class method:

from requests_oauth2client import OAuth2Client, ClientSecretJwt

oauth2client = OAuth2Client.from_discovery_endpoint(
    "https://url.to.the.as/.well-known/openid-configuration",
    auth=ClientSecretJwt("client_id", "client_secret"),
)

# OR, if you know the issuer value
oauth2client = OAuth2Client.from_discovery_endpoint(
    issuer="https://url.to.the.as",
    auth=ClientSecretJwt("client_id", "client_secret"),
)

This will fetch the document from the specified URI, decode it, and initialize an OAuth2Client pointing to the appropriate endpoint URIs.

If you use the issuer keyword argument, the URI to the discovery endpoint will be deduced from that identifier, and a check will be made to ensure that the issuer from the retrieved metadata document matches that value.

DPoP (Demonstrating Proof of Possession) is supported out-of-the-box. To obtain a DPoP token, you can either:

• pass dpop=True when using any OAuth2Client method that sends a token request,
• or enable DPoP by default by passing dpop_bound_access_tokens=True when initializing your client.

from requests_oauth2client import DPoPToken, OAuth2Client

oauth2client = OAuth2Client.from_discovery_endpoint(
    issuer="https://as.local",
    client_id="client_id", client_secret="client_secret",
)

token = oauth2client.client_credentials(scope="my_scope", dpop=True)
assert isinstance(token, DPoPToken)

# or, to enable DPoP by default for every token request
oauth2client = OAuth2Client.from_discovery_endpoint(
    issuer="https://as.local",
    client_id="client_id", client_secret="client_secret",
    dpop_bound_access_tokens=True,
)
token = oauth2client.client_credentials(scope="my_scope")
assert isinstance(token, DPoPToken)

DPoPToken is actually a BearerToken subclass. If you use it as a requests Auth Handler, it will take care of adding a DPoP proof to the request headers, in addition to the access token.

Since it is a BearerToken subclass, it is fully compatible with the requests compatible auth handlers provided by requests_oauth2client, such as OAuth2ClientCredentialsAuth, OAuth2AccessTokenAuth, etc. So you may use DPoP with those auth handlers like this:

import requests
from requests_oauth2client import OAuth2Client, OAuth2ClientCredentialsAuth, PrivateKeyJwt

client = OAuth2Client.from_discovery_endpoint(
    issuer="https://my.issuer.local",
    auth=PrivateKeyJwt("client_id", "client_secret"),
    dpop_bound_access_tokens=True, # enable DPoP by default
)

session = requests.Session()
session.auth = OAuth2ClientCredentialsAuth(
    client=client,
    scope="my_scope"
)

resp = session.get("https://my.api.local/endpoint")  # this will automatically obtain a DPoP token and use it
assert "DPoP" in resp.requests.headers  # the appropriate DPoP proof will be included in the request

Since DPoP is enabled by default with dpop_bound_access_tokens=True, then the OAuth2ClientCredentialsAuth will obtain and use DPoPToken instances. You could also leave it disabled by default and pass dpop=True when initializing you auth handler instance: OAuth2ClientCredentialsAuth(client=client, scope="my_scope", dpop=True).

By default, the private key used for signing DPoP proofs is auto-generated by OAuth2Client whenever a new token is obtained. By default, generated keys are of type Elliptic Curve (EC), and use the ES256 signature alg (as in Elliptic-Curve with a SHA256 hash). Should you, for testing purposes, wish to generate or use your own key, you may use the parameter dpop_key to provide a key of your choice. It takes a DPoPKey instance, which you can generate using DPoPKey.generate(), or by initializing an instance with a key that you previously generated:

from cryptography.hazmat.primitives.asymmetric import rsa
import jwskate
from requests_oauth2client import DPoPKey, DPoPToken, OAuth2Client

oauth2client = OAuth2Client.from_discovery_endpoint(
    issuer="https://as.local",
    client_id="client_id", client_secret="client_secret",
    dpop_bound_access_tokens=True,
)

dpop_key = DPoPKey.generate(alg="RS512")  # generate a new DPoP key with an alg of your choice
# or, for testing purposes only, your can load your own key
dpop_key = DPoPKey(private_key=jwskate.Jwk({"kty": "EC", "crv": "P-256", "alg": "ES256", "x": "...", "y": "...", "d": "..."}))
# or, any key material supported by `jwskate` is supported, so you can also use `cryptography` keys directly,
# but you need to specify the signature `alg` since it is not part of the key itself
dpop_key = DPoPKey(private_key=rsa.generate_private_key(public_exponent=65537, key_size=2048), alg="RS256")

token = oauth2client.client_credentials(scope="my_scope", dpop_key=dpop_key)
assert isinstance(token, DPoPToken)
assert token.dpop_key == dpop_key

Instead of generating your own keys everytime, you may also control how DPoPKeys are automatically generated. This can be useful for fuzz-testing, pen-testing or feature-testing the Authorization Server. To choose the signing alg, use the parameter dpop_alg when initializing your client. This will accordingly determine the key type to generate. You may also pass a custom dpop_key_generator, which is a callable that accepts a signature alg as parameter, and generates DPoPKey instances.

You can also override the DPoPToken class with a custom one, which will be used to represent the DPoP token that is returned by the AS, and then generates proofs and includes those proofs into HTTP requests.

You may use DPoPKey.generate as a helper method for that, or implement your own generator:

import secrets
from requests_oauth2client import DPoPKey, DPoPToken, OAuth2Client

class CustomDPoPToken(DPoPToken):
    """A custom DPoP token class that places the DPoP proof and token into a non-standard header."""
    AUTHORIZATION_HEADER = "X-Custom-Auth"
    DPOP_HEADER = "X-DPoP"

oauth2client = OAuth2Client.from_discovery_endpoint(
    issuer="https://as.local",
    client_id="client_id", client_secret="client_secret",
    dpop_bound_access_tokens=True,  # enable DPoP by default
    dpop_alg="RS256", # choose the signing alg to use, and it will automatically determine the key type to generate.
    dpop_key_generator=lambda alg: DPoPKey.generate(
        alg=alg,
        # those other parameters are for feature testing the AS, or for workarounding AS bugs:
        jwt_typ="jwt+custom", # you can customize the `typ` that is included in DPoP proof headers
        jti_generator=lambda: secrets.token_urlsafe(24), # generate unique jti differently than the default UUIDs
        iat_generator=lambda: 12532424, # override `iat` generation in DPoP proofs, here it will return a static value
        dpop_token_class=CustomDPoPToken, # override the class that represents DPoP tokens
    )
)

Authorization Server provided DPoP nonces are automatically and transparently handled by OAuth2Client.

Likewise, Resource Server provided DPoP nonces are supported when using the default DPoPToken class. This includes all requests-compatible auth handlers provided by requests_oauth2client, like OAuth2AccessTokenAuth, OAuth2ClientCredentialsAuth, OAuth2AuthorizationCodeAuth, etc.

As an example, see the sample below:

from requests_oauth2client import OAuth2Client, OAuth2ClientCredentialsAuth

import requests

oauth2client = OAuth2Client.from_discovery_endpoint(
    issuer="https://as.local",
    client_id="client_id", client_secret="client_secret",
)

response = requests.get(
    "https://my.api.local/endpoint",
    auth=OAuth2ClientCredentialsAuth(oauth2client, scope="my_scope", dpop=True),
)

Assuming that both the Authorization Server (at https://as.local) and the Resource Server (at https://my.api.local) require the use of DPoP nonces, then at least 4 different requests are sent as a result of the requests.get() call above:

1. The first request is to get a token from the Authorization Server, here using a Client Credentials grant and including a DPoP proof. DPoP also works with all other grant types. That first requests does not include a nonce. Since the AS requires a DPoP nonce, it replies to that request with an error=use_dpop_nonce flag and a generated DPoP nonce.

2. Second request is automatically sent to the AS, this time with a DPoP proof that contains the nonce provided by the AS. As a result, the AS returns a DPoP token.

3. Third request is sent to the target API, with the DPoP token obtained at step 2, and a DPoP proof that does not yet contain a nonce. The response from this call is a 401 with at least these 2 response headers:

   • a WWW-Authenticate: DPoP error="use_dpop_nonce" header, indicating that a DPoP nonce is requested,
   • and a DPoP-Nonce header containing the nonce to use.

4. a request is sent again to the target API, this time with a DPoP proof that contains the RS provided nonce obtained at step 3. Target API then should accept that request, do its own business and return a 200 response.

If you send multiple requests to the same API, instead of using individual calls to requests.get(), requests.post() etc., you should use a requests.Session or an ApiClient. It will make sure that the obtained access token and DPoP nonce(s) are reused as long as they are valid, which avoid repeating calls 1 and 2 unnecessarily and consuming more tokens and nonces than necessary:

from requests_oauth2client import ApiClient, OAuth2Client, OAuth2ClientCredentialsAuth

oauth2client = OAuth2Client.from_discovery_endpoint(
    issuer="https://as.local",
    client_id="client_id",
    client_secret="client_secret",
)

api = ApiClient("https://my.api.local/", auth=OAuth2ClientCredentialsAuth(oauth2client, scope="my_scope", dpop=True))
response1 = api.get("endpoint") # the first call will trigger requests 1. 2. 3. 4. like above
response2 = api.post("other_endpoint") # next calls will reuse the same token and DPoP nonces as long as they are valid.
# some time later
response3 = api.get("other_endpoint") # new tokens and DPoP nonces will automatically be obtained when the first ones are expired

AS and RS provided nonces are memoized independently by the DPoPToken instance, so the amount of "extra" requests to obtain new DPoP nonces should be minimal.

Using APIs usually involves multiple endpoints under the same root url, with a common authentication method. To make it easier, requests_oauth2client includes a requests.Session wrapper called ApiClient, which takes the root API url as parameter on initialization. You can then send requests to different endpoints by passing their relative path instead of the full url. ApiClient also accepts an auth parameter with an AuthHandler. You can pass any of the OAuth2 Auth Handler from this module, or any requests-compatible Authentication Handler. Which makes it very easy to call APIs that are protected with an OAuth2 Client Credentials Grant:

from requests_oauth2client import OAuth2Client, ApiClient, OAuth2ClientCredentialsAuth

oauth2client = OAuth2Client(
    "https://url.to.the/token_endpoint", client_id="client_id", client_secret="client_secret"
)
api = ApiClient(
    "https://myapi.local/root", auth=OAuth2ClientCredentialsAuth(oauth2client)
)

# will actually send a GET to https://myapi.local/root/resource/foo
resp = api.get("/resource/foo")

Note that ApiClient will never send requests "outside" its configured root url. The leading / in /resource above is optional. A leading / will not "reset" the url path to root, which means that you can also write the relative path without the / and it will automatically be included:

api.get("resource/foo")  # will also send a GET to https://myapi.local/root/resource/foo

You may also pass the path as an iterable of strings (or string-able objects), in which case they will be joined with a / and appended to the url path:

# will send a GET to https://myapi.local/root/resource/foo
api.get(["resource", "foo"])
# will send a GET to https://myapi.local/root/users/1234/details
api.get(["users", 1234, "details"])

You can also use a syntax based on __getattr__ or __getitem__:

api.resource.get()  # will send a GET to https://myapi.local/root/resource
api["my-resource"].get()  # will send a GET to https://myapi.local/root/my-resource

Both __getattr__ and __getitem__ return a new ApiClient initialised on the new base_url. So you can easily call multiple sub-resources on the same API this way:

from requests_oauth2client import ApiClient

api = ApiClient("https://myapi.local")
users_api = api.users
user = users_api.get("userid")  # GET https://myapi.local/users/userid
other_user = users_api.get("other_userid")  # GET https://myapi.local/users/other_userid
resources_api = api.resources
resources = resources_api.get()  # GET https://myapi.local/resources

ApiClient will, by default, raise exceptions whenever a request returns an error status. You can disable that by passing raise_for_status=False when initializing your ApiClient:

from requests_oauth2client import ApiClient

api = ApiClient(
    "http://httpstat.us", raise_for_status=False
)  # raise_for_status defaults to True
resp = api.get("500")
assert resp is not None
# without raise_for_status=False, a requests.exceptions.HTTPError exception would be raised instead

You may override this at request time:

# raise_for_status at request-time overrides the value defined at init-time
resp = api.get("500", raise_for_status=True)

You can access the underlying requests.Session with the session attribute, and you can provide an already existing and configured Session instance at init time:

import requests
from requests_oauth2client import ApiClient

session = requests.Session()
session.proxies = {"https": "http://localhost:3128"}
api = ApiClient("https://myapi.local/resource", session=session)
assert api.session == session

requests_oauth2client is flexible enough to handle most use cases, so you should be able to use any AS by any vendor as long as it supports OAuth 2.0.

You can however create a subclass of OAuth2Client or ApiClient to make it easier to use with specific Authorization Servers or APIs. OAuth2Client has several extensibility points in the form of methods like OAuth2Client.parse_token_response(), OAuth2Client.on_token_error() that implement response parsing, error handling, etc.

from requests_oauth2client.vendor_specific import Auth0

a0client = Auth0.client(
    "mytenant.eu", client_id="client_id", client_secret="client_secret"
)
# this will automatically initialize the token endpoint to https://mytenant.eu.auth0.com/oauth/token
# and other endpoints accordingly
token = a0client.client_credentials(audience="audience")

# this is a wrapper around Auth0 Management API
a0mgmt = Auth0.management_api_client(
    "mytenant.eu", client_id="client_id", client_secret="client_secret"
)
myusers = a0mgmt.get("users")