A delegation chain is the ordered sequence of authority transfers from an original principal, through any intermediate services or agents, to the actor that finally performs an action. Each link records who delegated what to whom; the chain as a whole is what makes the final action attributable.
Why it matters
A single delegation hop is easy to reason about: a principal grants a scoped right to one agent. Real systems are rarely that shallow. An agent calls another agent, which calls a service, which calls an API — each hop acting for the one above. Without a chain, the resource at the end sees only the last credential and cannot tell which human or organisation stands behind it. Without a chain, you cannot answer the KYA question — on whose authority did this happen? — after the fact.
How it works in the standards
OAuth is the lineage. RFC 8693 (OAuth 2.0 Token Exchange) lets a service present a token and receive a new one for a different subject or scope, with explicit semantics for delegation versus impersonation. Delegation keeps both identities visible — the principal and the acting party — so a chain can be reconstructed; impersonation collapses them and hides the trail. RFC 8693 defines an `act` (actor) claim precisely so each hop can name who is acting on whose behalf.
OAuth 2.1 consolidates the core framework and its security guidance, and is the baseline the agent-specific work extends. The IETF drafts for AI-agent authentication and on-behalf-of-user flows extend OAuth so the agent’s identity is carried through consent and into the token, and the OpenID Foundation’s Identity Management for Agentic AI examines the harder case: recursive delegation and cross-domain trust propagation that current frameworks were not built for.
What each link should carry
- The delegating party and the receiving party.
- The permitted actions, resources, and any amount or data limits.
- A validity window and expiry.
- A revocation handle.
- A reference to the previous link, so the chain is verifiable end to end.
How it works in practice
Token exchange turns these into auditable hops: rather than each service overwriting the last credential, each obtains a fresh token that names the original subject and the current actor. That is what prevents the confused deputy problem, where a privileged intermediary acts without checking who asked. When a chain is shortened to impersonation for legacy reasons, log it and scope it tightly. See What is Know Your Agent? and the standards tracker.
Related terms
Sources
- IETF, RFC 8693, OAuth 2.0 Token Exchange (Accessed 2026-10-08)
- IETF, RFC 6749, The OAuth 2.0 Authorization Framework (Accessed 2026-10-08)
- IETF, draft-ietf-oauth-v2-1, The OAuth 2.1 Authorization Framework (Accessed 2026-10-08)
- IETF, draft-oauth-ai-agents-on-behalf-of-user (Accessed 2026-10-08)
- OpenID Foundation, Identity Management for Agentic AI (Accessed 2026-10-08)
Know Your Agent (KYA) explains agent identity, verification and accountability. This is an explainer, not legal or compliance advice — see our Sources & methodology. New to KYA? Start with What is Know Your Agent? and the glossary.