I'd like to get feedback on a design and a working prototype for a long-standing
request: restricting which users may use which clients, based on roles, group
membership or user attributes, for a single client or a set of clients
("application groups"). Today people solve this with a conditional subflow plus
the Deny Access authenticator, or with community extensions such as
keycloak-restrict-client-auth. Both approaches stop at the browser flow: they do
not cover token refresh, direct grant or client credentials, and per-client flow
overrides do not scale to sets of clients.
The prototype lives on my fork: