Google ADK

Google ADK keeps its own auth hook. Authdog stays the identity provider. The planned sample attaches an agent access token there and refuses the run when that token is missing or userinfo rejects it.

Sample planned at adk/python in authdog/agents-samples. That folder is not published yet, so this page does not pin google-adk. Python 3.10+.

Channel auth, machine credentials, tool checks, and instructions for this framework are the other pages under Google ADK in AI.

The hook

def before_agent_callback(callback_context):
    token = callback_context.state.get("authdog_token")
    subject = subject_if_accepted(token)
    callback_context.state["authdog_token"] = None
    if subject is None:
        return types.Content(
            role="model",
            parts=[types.Part(text="unauthorized")],
        )
    callback_context.state["subject"] = subject
    return None

The parameter name is callback_context. ADK passes callback arguments by keyword. Returning types.Content skips the agent body. Returning None lets it run. The token lives in callback state, which the host set. It is not the user message. Clear it before the model runs. AuthCredential on a tool is outbound tool auth. It is not this check, and it is not a place for the agent client secret.

This is agent identity: client_credentials, then this hook. It is not Agent Sign-In. A browser sign-in does not replace the hook.

What you should see

A callback with no token, or a token userinfo rejects, returns content whose text is unauthorized. The agent body does not run.

A callback with a bearer userinfo accepts stores the subject and returns None.

pytest in adk/python is how the sample will prove both. It does not call a model.

Configure

The shared exchange is in Agent frameworks. There is no refresh token. Repeat it before expires_in.

Related