AI agents are calling your APIs. Your API gateway sees the request, your authorization server issues the token. But neither of them knows which agent made the call, whether it’s the same agent as last time, or whether it’s authorized to act on behalf of the user it claims to represent.

This session is about the identity layer that agentic systems are currently missing, and the emerging standard that addresses it: draft-ietf-oauth-client-id-metadata-document. We’ll cover why traditional OAuth client registration breaks down at agent scale, how the spec enables dynamic client identity without a registry, and why knowing the URL proves nothing, possession of the private key does.
We’ll also cover what the spec doesn’t solve on its own, instance-level identity, the bootstrapping problem, and the gap between where the standard is and where enterprise authorization server support currently sits.

The goal is to leave you better equipped to evaluate, design, and challenge the identity layer in any agentic system you are building or buying.