Skip to main content
Version: Next

Check whether a User has Access to Something

This guide will explain how you can use Ory Keto's check-API to determine whether a subject has a specific relation on an object. The result can be used for controlling access to specific resources.

Synchronous Authorization Flow#

We recommend offloading the whole burden of access control to Ory Keto. Typically, this means that the application forwards every incoming request as a check request to Ory Keto. The following chart demonstrates how such a flow can look like:

Note that the channel of communication between user <-> application, and application <-> Ory Keto can vastly differ. The application could offer a JSON API towards the user, while communicating with Keto through gRPC.

As a first step, the application has to authenticate the user reliably to provide the subject to Keto. This can be achieved e.g. by using Ory Kratos or any other authentication system.

The request (here decypher of the message 02y_15_4w350m3) is then translated into a request to Ory Keto's check-API. Basically, the application is asking Keto "Is john allowed to decypher the text 02y_15_4w350m3?"

This question is encoded as the following relation tuple:


It is up to the application and its defined relation tuples how the check requests have to be encoded. In this example we assume that the known cypher messages are stored in Ory Keto and access to the cleartext is encoded by the decypher relation.

Directly Defined Access#

Ory Keto can know the exact relation tuple that the application is checking. Intuitively, this means that john was allowed to decypher the message 02y_15_4w350m3 directly (imagine a "Share with john" input in a UI).

Try this yourself by first adding the relation tuple using the write API:


Now, we can use the check-API to verify that john is allowed to decypher the message:


Indirectly Defined Access#

On the other hand, it is possible to indirectly grant john access to the resource. This could e.g. be done by adding a group, lets call it hackers. Now we can grant access to the resource to everyone in that group by adding the following relation tuple to Ory Keto:


We also have to make john a member of hackers by adding the relation tuple:


Now, when Keto receives above check request, it will resolve the subject set


and determine that john is a subject in the resulting set. Therefore, it approves the check request.

There is no limit on the number of indirections through subject sets. It is however important to follow our best practices to ensure a good performance.

Caching Keto's responses#

We do not recommend that you cache the responses from Ory Keto. It is designed to respond quickly and still provide some consistency guarantees. Especially, for the revocation of access it is important to not use a local cache. Be ensured that Ory Keto heavily utilizes caching wherever possible. If you still happen to find unacceptably slow check requests, check that you follow our best practices for good performance, or open an issue if the problem still persists.


We learned how to integrate check requests and access control into an application using Ory Keto's check-API.

Last updated on by aeneasr