Dynamic client registration lets a software client, such as an AI agent, register itself with an authorization server at runtime rather than requiring an administrator to manually provision it in advance. This matters for agentic identity because many AI agents are chosen and configured by the end customer, not pre-registered by the business whose systems the agent is accessing.
Frequently asked questions
Why can't every agent just be manually pre-registered?
Customers increasingly choose and configure their own agents, so requiring manual provisioning for each one doesn't scale and would block legitimate new agents from ever connecting.
What's the alternative to dynamic client registration for identifying new clients?
Emerging approaches like Client ID Metadata Documents let an authorization server validate an unfamiliar client without either manual registration or fully open dynamic registration. See CIMD.
.avif)