]> Google, LLC

nwatson@google.com

Security Web Authorization Protocol oauth refresh token authorization token endpoint This specification extends OAuth 2.0 by adding new token endpoint response parameters to specify refresh token expiration and user authorization expiration. About This Document The latest revision of this draft can be found at . Status information for this document may be found at . Discussion of this document takes place on the Web Authorization Protocol Working Group mailing list (), which is archived at . Subscribe at . Source for this draft and an issue tracker can be found at .

Introduction RFC6749 defines the OAuth 2.0 protocol, part of which is the ability for a client to receive a refresh token that may be repeatedly exchanged for more access tokens. OAuth 2.0 does not contain any normative language around expiration or lack thereof for refresh tokens, mentioning only that they are "typically long-lasting". In the years since the publication of OAuth 2.0, in response to changing security and privacy landscapes, many authorization servers have begun to issue shorter-lived refresh tokens for two main reasons:

• The authorization server or user may decide that the access being granted is too sensitive to allow indefinite access (e.g. mail or health data).
• The authorization server enforces a maximum duration that refresh tokens may be held without being exchanged on the token endpoint.

Clients may wish to implement special handling for expiring refresh tokens. For example, if the user has granted expiring access, the client may notify the user that they will need to reauthorize access before a certain date to avoid interruption of service.

Requirements Notation and Conventions The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

Terminology This specification uses terminology defined in . The following terms are used throughout this document:

Resource owner and user

May be used interchangeably to refer to the entity capable of granting access to a protected resource.

Client, application, and relying party

May be used interchangeably to refer to the application making protected resource requests on behalf of the resource owner and with its authorization.

Authorization

The resource owner's permission grant for a client to access protected resources on their behalf, as described in Sec 4.1.1.

Access token

A credential used by the client to access protected resources on behalf of the resource owner, as referenced in Sec 1.4. Access tokens represent proof of authorization.

Refresh token

A credential used by the client to obtain new access tokens without prompting the user, as referenced in Sec 1.5. Refresh tokens do not grant authorization or renew authorization, they only provide a mechanism for obtaining new access tokens within the bounds of an existing authorization.

Concepts There are two mechanisms that can affect refresh token expiration.

Authorization expiration When granting authorization for an application to access their data as referenced in Sec 4.1.1, the user may opt to time-limit that authorization, especially if the data is sensitive or they aren't sure how long they'll continue using the application. The authorization server itself may also impose mandatory limits on authorization duration.

Refresh token timeout Authorization servers may wish to define a maximum amount of time clients can hold a refresh token without exchanging it. Beyond the security benefit provided by expiring credentials, this also provides a convenient mechanism for authorization servers to ensure there aren't ancient valid credentials out in the wild, which could complicate tasks like refresh token key rotation.

Refresh token expiration The refresh token MUST NOT expire later than the user authorization expires. It MAY expire earlier if the authorization server also enforces a maximum duration between refresh token exchanges. If the user renews their authorization, the authorization server SHOULD update the expiration time of existing refresh tokens if their lifetime was truncated due to user authorization expiration. (This is especially true if the authorization was updated out of band as discussed in User Experience Considerations.) The authorization server MUST NOT accept expired refresh tokens for any purpose, even if it has no way to update the expiration time of existing refresh tokens. Access tokens MUST NOT expire later than the user authorization expires. If the user renews their authorization, the authorization server MAY update the expiration time of existing access tokens if possible. Resource servers MUST NOT accept expired access tokens for any purpose, even if the authorization server has no way to update the expiration time of existing access tokens.

Token endpoint response This specification introduces two new response parameters.

Successful response An authorization server MAY return only one of these parameters. Providing refresh_token_timeout without authorization_expires_in indicates that the user's authorization is indefinite, but the refresh token must be used within the specified timeout to remain valid. Providing authorization_expires_in without refresh_token_timeout indicates that the authorization has a fixed duration, but the refresh token has no maximum idle time and may remain valid if the authorization is extended out of band. If finite, the authorization server MUST return these values whenever the token endpoint response contains the refresh_token field. The authorization server MAY return these values even if the response contains no refresh_token field, in which case the values correspond to the presented refresh_token. This can be useful in the following example cases:

• For refresh_token_timeout, the authorization server could have updated the existing refresh token lifetime in place.
• For authorization_expires_in, the user's authorization lifetime could have been modified out of band.
• In all cases, it can be convenient for the client to receive these values in each response.

Relationship of authorization_expires_in to scopes Though authorization_expires_in is returned from the token endpoint when refresh tokens are used, it corresponds to the user's authorization for scopes (or finer-grained access through RAR ) rather than individual tokens. The authorization server SHOULD ensure consistent lifetimes across multiple refresh tokens for the same scopes. Tying authorization lifetime to scopes means it's possible to have some access valid for one duration and other access valid for a different duration. For example, a user could grant indefinite access for the openid scope and short-lived access for a calendar scope. In situations like this, it is RECOMMENDED that the authorization server return the minimum time that any access granted by the refresh token is valid. This does run some risk of the client asking the user to reauthorize prematurely. In the previous example, the client might ask the user to reauthorize the openid scope because it received an authorization_expires_in value corresponding to the short-lived calendar scope. If clients are requesting multiple scopes that can have different lifetimes, they will ultimately need to make their own tradeoffs to decide how and when to ask the user for reauthorization. This specification's goal is simply to provide them with more information to make this decision.

Indefinite Expiration Omitted values indicate that there is no fixed upper bound on the lifetime of the credential or authorization. If the authorization server has not declared its support for refresh token lifetime in the Authorization Server Metadata, omitted response fields could indicate either indefinite validity or simply lack of support for this specification. However, indefinite expiration and lack of information about expiration should be handled by the client in the same way. That is to say, the client must always handle refresh token invalidation not caused by expiration, such as by explicit user revocation. Clients MUST NOT make any assumptions that omitted response fields in one response imply their omission in later responses too. Rather than omitting a response value, an authorization server may choose to return a large arbitrary value, e.g. 315569520 for 10 years. This avoids any ambiguity around support for indefinite values while achieving a similar practical effect. Clients MUST treat all large values as literals and MUST NOT make any assumptions about which may be considered indefinite.

Error response The existing invalid_grant error code already explicitly covers token expiration and should be sufficient. Upon receiving this error code the client SHOULD start a new authorization grant flow.

Example Suppose an authorization server enforces that refresh tokens must be exchanged at least once every 7 days, and a user has granted authorization to an application for access for 10 days. The initial authorization code grant (Day 0) will result in the following response values: A refresh token grant on Day 2 will result in the following response values: A refresh token grant on Day 7 will result in the following response values: If instead, the client held the initial refresh token for 8 days (i.e. exceeding refresh_token_timeout but not authorization_expires_in), the refresh token grant will fail: Note that the error description text is non-normative and for illustrative purposes only.

Update to Token Introspection While Token Introspection is primarily intended for resource servers seeking information about received access tokens, the specification does permit refresh token introspection as well. Authorization servers supporting refresh token introspection SHOULD support the refresh_token_timeout and authorization_expires_in parameters on the endpoint with the same semantics as defined in Token endpoint response. These parameters SHOULD be returned only when the presented token is a refresh token. Use of a refresh token on token introspection MUST NOT reset any refresh_token_timeout duration.

Update to Authorization Server Metadata Support for the expiring refresh tokens SHOULD be declared in the OAuth 2.0 Authorization Server Metadata with the following metadata: If the authorization server omits expiration time response fields to indicate indefinite validity, it MUST declare refresh_token_expiration_types_supported in its metadata to indicate to the client that it's aware of this spec.

User Experience Considerations While clients must be able to gracefully handle tokens' expiring at any time, the user experience may suffer if there's an unintended interruption of service. This degradation of experience would most likely be felt by users of clients running in the background, such as task or travel management apps that rely on access to a user's calendar or inbox. If an application recognizes that its access is nearing expiration, it can proactively prompt the user for reauthorization next time they're "in the loop" (e.g. using a parameter like prompt=consent from ), or even communicate to the user out of band that their granted access is expiring. Another option an authorization server could provide to the user is a management surface where the user can go proactively extend the lifetime of their own grant, which would update the lifetime of the client's refresh token(s) in place. The client would discover the extended expiration on its next refresh token grant request.

Security Considerations While it is possible to allow refresh token expiration to exceed that of user authorization expiration if the authorization server checks both timestamps when validating a refresh token, this is a potentially dangerous source of bugs in systems with complicated user authorization models. By requiring refresh tokens to expire no later than user authorization expires, there is less risk of bugs that accidentally provide data access to the client beyond the term of the user's authorization. Authorization servers implementing token rotation on every refresh [RFC 9700] Sec 4.14 may wish to enforce a maximum duration that a refresh token may be held without rotation, and this specification allows that duration to be communicated as part of the API rather than relying on documentation. Clients may wish to maintain multiple refresh tokens with different access in order to separate different lifetimes across different scopes. For example, a short-lived token to access financial data and a long-lived token to access basic user info. There is a tradeoff here, both in complexity of token management and also in increased friction for the user to authorize multiple tokens.

Privacy Considerations Allowing users to time-limit their authorization is a privacy improvement. While this was already doable in regular OAuth implementations, the potential interruption of service for the user may have discouraged implementation of the feature. This specification provides a standardized way to mitigate that concern and should lead to greater adoption of time-limited authorization.

IANA Considerations

OAuth Parameters Registration This specification registers the following OAuth parameter definitions in the IANA OAuth Parameters registry.

Registry Contents

• Name: refresh_token_timeout
  • Parameter Usage Location: token response
  • Change Controller: IETF
  • Reference: This document
• Name: authorization_expires_in
  • Parameter Usage Location: token response
  • Change Controller: IETF
  • Reference: This document

OAuth Authorization Server Metadata Registration This specification registers the following Authorization Server Metadata definitions in the IANA OAuth Authorization Server Metadata registry.

Registry Contents

• Metadata Name: refresh_token_expiration_types_supported
  • Metadata Description: What types of refresh token expiration are supported by the authorization server
  • Change Controller: IETF
  • Reference: This document

Change History Delete this section before publication.

• May 8, 2026:
  • Incorporate Vanshaj's review:
    • Add language on backwards compatibility for definite duration becoming finite.
    • refresh_token_expiration_types_supported "credential" value renamed to "token_timeout".
    • Simplified example
    • Linking UX considerations from RT expiration section
• Feb 27, 2026:
  • Address Issues 4, 5, 6 from George to discuss tradeoffs around managing multiple tokens or scopes with different expirations, as well as out of band reauthorization by the user.
  • Rewording and clarification based on Dan's suggestions on the list.

References Normative References The OAuth 2.0 authorization framework enables a third-party application to obtain limited access to an HTTP service, either on behalf of a resource owner by orchestrating an approval interaction between the resource owner and the HTTP service, or by allowing the third-party application to obtain access on its own behalf. This specification replaces and obsoletes the OAuth 1.0 protocol described in RFC 5849. [STANDARDS-TRACK] This specification defines a method for a protected resource to query an OAuth 2.0 authorization server to determine the active state of an OAuth 2.0 token and to determine meta-information about this token. OAuth 2.0 deployments can use this method to convey information about the authorization context of the token from the authorization server to the protected resource. This specification defines a metadata format that an OAuth 2.0 client can use to obtain the information needed to interact with an OAuth 2.0 authorization server, including its endpoint locations and authorization server capabilities. This document describes best current security practice for OAuth 2.0. It updates and extends the threat model and security advice given in RFCs 6749, 6750, and 6819 to incorporate practical experiences gathered since OAuth 2.0 was published and covers new threats relevant due to the broader application of OAuth 2.0. Further, it deprecates some modes of operation that are deemed less secure or even insecure. In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. Informative References This document specifies a new parameter authorization_details that is used to carry fine-grained authorization data in OAuth messages.

Acknowledgments A special thanks to Aaron Parecki for his continuous feedback and invaluable assistance in guiding the author through the IETF draft process. The author would also like to thank Andrii Deinega, George Fletcher, Dan Moore, and Vanshaj Singhania for their reviews and technical suggestions on the repo and mailing list.