Skip to content
IAM Insights

Explainers · Glossary

OAuth Identity and Authorization Chaining Across Domains

Identity Chaining · draft-ietf-oauth-identity-chaining · OAuth Identity Chaining · identity and authorization chaining

By Boston Identity · Explained · 6 min read

The short answer

A way to carry who a user is, and what they allowed, from one organization's OAuth system into another's. The home server issues a signed grant; the far server swaps it for its own access token.

What it is
A protocol
Defined by
IETF OAuth working group draft (-17), approved and waiting at the RFC Editor
Who it's for
Teams whose agents or services call APIs in another organization
In our briefings
28By month of the source: 2025-12: 0, 2026-01: 0, 2026-02: 1, 2026-03: 1, 2026-04: 0, 2026-05: 2, 2026-06: 4, 2026-07: 6, 2026-08: 1, 2026-09: 5, 2026-10: 6, 2026-11: 00by month, last 12
01

What is OAuth Identity and Authorization Chaining Across Domains?

OAuth (the open standard for delegated access to APIs) works well inside one trust domain, the set of systems that answer to one authorization server. A request often has to go further. Every service on the path needs to know which user started the request and what that user allowed.§1 Identity and Authorization Chaining Across Domains is the Internet Engineering Task Force (IETF) draft that keeps that information intact when a request crosses into another organization's systems.§1

It defines no new token. It combines two published standards: token exchange (RFC 8693), done at home, and the JSON Web Token (JWT) bearer grant (RFC 7523), used at the far side.§2 A home system may hand out the grant through its own interface instead of token exchange.§2 Each crossing repeats the same steps, so the pattern can run through several domains in turn.S1 It is a draft, not yet a standard. The datatracker showed revision 17 approved and waiting only for the RFC Editor.1

Not the same as plain token exchange, which trades one token for another at a single authorization server. Chaining normally uses that exchange at home only to get a grant that a different organization's server will accept.

02

How does OAuth Identity and Authorization Chaining Across Domains work?

Four parties take part. A client in domain A holds a token from its own authorization server, and it wants a resource behind domain B's authorization server. Domain B already trusts domain A's signing keys.§2.1

Figure steps
  1. The client finds domain B's server
  2. The client asks its home server for a grant
  3. Domain A's server issues a signed grant
  4. The client presents the grant to domain B
  5. Domain B's server checks it and issues a token
  6. The client calls the resource
Fig. 1

The client trades its domain A token for a signed grant at home, swaps that grant for domain B's access token, then calls the resource.

Redrawn from the draft's Figure 1

Swipe for the full figure →

  1. The client finds domain B's server

    The client works out which authorization server guards the resource in domain B. It may read the resource's published metadata (RFC 9728), keep a fixed mapping, or use other means.§2.2

  2. The client asks its home server for a grant

    The client sends the token it already holds to domain A's authorization server as an RFC 8693 token exchange. The request must name domain B's server, by its address or by a logical name. The client may also ask for a narrower scope.§2.3.1

  3. Domain A's server issues a signed grant

    Domain A's server must refuse if the target is unknown or its policy says no.§2.3.2 Otherwise it returns a JWT authorization grant, signed with its own key.§2.1 The grant's audience, the server it is meant for, must name domain B's server. The draft recommends naming only one, so domain B cannot pass the grant on to a third domain.§2.3.3

  4. The client presents the grant to domain B

    The client sends the grant to domain B's token endpoint as a JWT bearer assertion under RFC 7523.§2.4.1 It can say which resource it wants with the scope or resource parameter.

  5. Domain B's server checks it and issues a token

    Domain B's server validates the grant's signature and audience against the keys it trusts for domain A.§2.1 It must deny the request if it cannot identify the subject, the user the grant is about. It may also deny it on policy, such as no trust relationship with domain A.§2.4.2 If all is well, it returns its own access token.§2.4.3

  6. The client calls the resource

    The client uses domain B's access token to call the protected resource in domain B.§2.1

03

Why does OAuth Identity and Authorization Chaining Across Domains matter now?

Agents now cross company lines. A vendor's agent may act on a customer's data, and other agent work points at this draft for that case. The transaction-tokens draft keeps its tokens inside one trust domain and hands the crossing to identity chaining.2 The six-vendor agent draft names it for agent parts that sit behind different authorization servers.1 The workload identity architecture (WIMSE) also sends a workload out of its home domain by this pattern. Its alignment section there still reads as a to-do.3

Of the standards in that conversation, this one is nearest to done. We forecast that the RFC Editor publishes it with an RFC number by 31 March 2027.4 Okta and Ping Identity each have an author on it.1

The crossing is where a customer can narrow a vendor's agent. Either server may add, remove or remap claims. A client may ask for less, and the server checks that the scopes it asks for are no wider than the token it presented.5 A newer profile for transaction tokens copies this design. Each server keeps its output no wider than its input, one crossing at a time.6

04

Where does it fit?

Compared withHow OAuth Identity and Authorization Chaining Across Domains differs
OAuth Token ExchangeToken exchange is one step at one server. Chaining uses it at the home server to get a grant aimed at a foreign server, then adds a second step there.
JWT Bearer GrantThe JWT bearer grant is how domain B accepts a signed assertion. The chaining draft describes getting that assertion by token exchange at domain A's server, and it does not rule out other ways.
Transaction TokensTransaction tokens carry context between services inside one trust domain and stop at its edge. Chaining is the sibling draft that handles the crossing between domains.
Identity Assertion Authorization GrantID-JAG (the identity assertion JWT authorization grant) is a profile that starts from a person's single sign-on assertion at an identity provider both sides trust. It and a newer transaction-token profile are sibling grants built on the chaining framework.7
05

What goes wrong?

  • The grant is a bearer token, so anyone who steals it can replay it at domain B. The draft lists short lifetimes, single use and client authentication as mitigations, and domain B's server can enforce them.§5.5
  • Authorization servers should sender-constrain tokens as the OAuth security best practice (RFC 9700) advises. The draft admits that bearer tokens are still the most widely deployed kind.§5.2
  • A stolen home token could become a lateral move into domain B. Domain A's server can check that the client is allowed to present that token before issuing a grant. The draft recommends this check where it can be run.§5.3
  • Domain B should not issue refresh tokens to the client. A refresh token would let the client keep getting access after the user logs out or loses access in domain A.§5.4
  • The draft does not define the format of transcribed claims.§2.5 One proposal offers a delegation chain for that gap, but it gives two accounts of who signs the earlier records.8
  • The draft carries which resource servers were called earlier only as an option.§2.5 Beyond that option, nothing in the token says who acted in between, and nothing extends per hop.9 One implementer found the draft silent on a subject token with no subject claim.10
06

What should a security leader ask?

  1. 01

    Which of our partners' authorization servers trust our signing keys today, and who approved each one?

  2. 02

    When a vendor's agent reaches our data, does our authorization server narrow its scope and claims, or pass everything through?

  3. 03

    Do the grants we issue name a single audience and expire within minutes, and can each be used only once?

  4. 04

    Does any partner domain issue refresh tokens on a chained grant, so access outlives the user's session with us?

  5. 05

    If a request crosses three domains, how would we trace which service acted at each hop?

07

Sources

›Evidence: 10 claims from 6 briefings
  1. 1BI-FIND-2026-0184 · The cross-domain OAuth draft is one step from an RFC, and it never mentions agentsOf every standard the agent-identity conversation points at, this is the one nearest to done. The datatracker shows draft-ietf-oauth-identity-chaining at revision 17, Standards Track, IESG state 'RFC Ed Queue': approved, waiting only for the RFC Editor. The authors are Arndt Schwenkschuster and Pieter Kasselman of Defakto, Kelley Burgin of MITRE, Mike Jenkins of the NSA's commercial cybersecurity arm, Brian Campbell of Ping Identity and Aaron Parecki of Okta, and the consensus is tighter than the list suggests: Campbell co-wrote both RFCs this draft combines, and Parecki co-wrote the discovery RFC it uses to find the other domain. The transaction-tokens draft names it as the cross-domain mechanism, and the six-vendor agent draft names it for the case where an agent's components sit behind different authorization servers. Both of those are further from publication than this.
  2. 2BI-FIND-2026-0174 · OAuth's transaction token goes unchanged down a service chain and names who started it, not who acted in betweenA transaction token is valid inside one trust domain and stops at its edge. Crossing into another domain is handed to a sibling draft, OAuth Identity and Authorization Chaining Across Domains, which Kasselman also co-authors, so the two are designed to interlock: transaction tokens inside, identity chaining between. For the case the summit's Aembit CEO called the hard half, an agent that runs inside a SaaS vendor's product and acts on a customer's data, this draft therefore offers nothing directly. The vendor's domain and the customer's are two trust domains, and the token that would carry the customer's user and authorization context across that boundary is the chaining draft's job, which nobody in our corpus has yet read.
  3. 3BI-FIND-2026-0300 · The IETF's workload-identity architecture gives AI agents one subsection, calling an agent a delegated workloadA WIMSE identity is local by design, and the way out of the local domain is unfinished. Identifiers are scoped to a trust domain, and credentials 'constrain the subjects and actors identified in delegation and impersonation tokens to be bound by a trust domain'. To reach another organization a workload trades its credential at a token service for one the other side's token service will accept, following the OAuth identity-chaining pattern, and the section that should align the two still reads '(TODO: align with OAuth cross domain identity and authorization)'. An agent hosted by one company and acting inside another is the case this leaves open.
  4. 4BI-CALL-2026-0055 · The cross-domain OAuth draft is one step from an RFC, and it never mentions agentsBy 31 March 2027 the RFC Editor publishes OAuth Identity and Authorization Chaining Across Domains as an RFC with a number. The draft is already past the IESG and in the editor's queue, and the queue's normal residence is a few months. Wrong if on that date the datatracker still shows an IESG state rather than an RFC number, or the document has been returned to the working group.
  5. 5BI-FIND-2026-0186 · The cross-domain OAuth draft is one step from an RFC, and it never mentions agentsThe interesting work happens at the crossing, and the draft gives the receiving side the pen. Authorization servers on either side of the hop may add, remove or remap claims: a subject identifier that differs between domains is rewritten; a bank's server strips 'account type: premium' and passes only a transaction limit, so 'domain B receives only the claims relevant to its operations'; and a client may ask for less, with the server checking that 'the requested scopes are not higher privileged than the scopes of the presented subject_token'. That is the transaction-tokens draft's never-wider rule applied across a boundary, and it is where a customer's own authorization server would narrow a SaaS vendor's agent before it touched anything. The draft then leaves the claim format and the transcription mechanics undefined, which is the gap any implementation will fill on its own terms.
  6. 6BI-FIND-2026-0494 · A cross-domain transaction-token grant names who started the transaction and drops the agent that askedNarrowing is a no-wider rule that authorization servers check one crossing at a time. Domain A's server "MUST NOT" issue a grant broader than the Txn-Token's scope (the grant's scope is the intersection of the token's and the request's), and domain B's server must independently keep its access token within the grant. Applied across three or more domains, each crossing is self-contained under its own bilateral trust agreement: nothing ties the third hop back to the first, the draft concedes a grant's claims may derive from what an upstream domain sent, and only scope strings narrow, with rich authorization details left to a later revision.
  7. 7BI-FIND-2026-0496 · A cross-domain transaction-token grant names who started the transaction and drops the agent that askedIf adopted, OAuth's cross-domain toolkit would carry two sibling JWT grants built on the same chaining framework. ID-JAG starts from a person's single sign-on assertion at an identity provider both sides trust; this one starts from a Txn-Token, covers people, systems and workloads, and rests on a bilateral agreement instead of a shared provider, with no client_id, tenant claims or authorization details. A server running both tells them apart by the typ header. The OAuth list already has a thread titled after a direct ID-JAG-to-Txn-Token exchange (Issue #17), a step this -00 does not contain.
  8. 8BI-FIND-2026-1289 · A June OAuth draft logs each hand-off between AI agents, but the per-hop limits the act claim lacks are optionalAcross trust domains, the draft offers delegation_chain as the form for the claims transcription that the OAuth working group's identity-chaining draft leaves undefined, and the receiving server MUST copy the chain into the token it issues. It then gives two accounts of who signs the upstream records. Section 8.3 has resource servers verify each record with the key of the server that issued it, which "may differ across records"; section 10.7 says the receiving server re-signs the upstream records, so every record verifies with the current issuer's key. We read these as two different tokens, and the draft does not say which a resource server should expect.
  9. 9BI-FIND-2026-0187 · The cross-domain OAuth draft is one step from an RFC, and it never mentions agentsThe draft's appendix answers part of the question Uber left open, for the cross-domain case and without saying 'agent'. A resource server may act as the client, discovering the other domain's authorization server from that domain's own protected-resource metadata; and an authorization server may act as the client for callers that cannot reach the other domain themselves, so that 'an authorization server in trust domain A may obtain an access token from an authorization server in trust domain B on-behalf-of any client in trust domain A'. Delegated key binding carries the original client's key across in a requested_cnf parameter so the resulting token stays bound to it. That is a standards-track description of an intermediary carrying authority across a boundary on a client's behalf, which is what an agent gateway is sold as. What it does not carry is an actor chain: nothing in the token says who acted in between, and nothing extends per hop. Medium, because the relevance to agents is ours and the draft's own use cases are CI pipelines, partner dashboards and cross-domain calendars.
  10. 10BI-FIND-2026-0894 · One implementer found a transaction-token rule that let any scope through; the OAuth group fixed it in weeksThe identity-chaining draft, already approved and waiting at the RFC Editor, is silent on a subject token that carries no sub claim. Agnihotri's server issues the chaining grant anyway and refuses it only when it is presented, which he calls allow and defer, and he proposes text letting a server either synthesise a subject or refuse, with refusal recommended. Both gaps in the report come from one implementation with no interoperability testing, and the author says so.

›In our briefings

The claim each of the newest briefings on OAuth Identity and Authorization Chaining Across Domains would have you leave with.

  1. Revision -02 of Shankarlingam's draft says the delay before a disabled agent is refused is "a property of the receiver's configuration, not of the token alone", and that passing the receiver's checks does not show a disabled client is refused within the bound the receiver computes. Raising the maximum token age, DNS caching, caching of issuer metadata and keys, or the clock-skew allowance lengthens the time a disabled client is still accepted, while every signature, issuer, client, audience and DNS check keeps passing. Those checks, the draft says, "establish that a token is authentic; they do not establish that the client is still enabled." A receiver SHOULD recompute the bound before any change to those settings takes effect, and SHOULD test it directly, again after any change to those settings: obtain link tokens for a test client, disable it at a time not known to whoever observes the receiver, keep presenting tokens issued before the disable, and compare the last acceptance with the time the authorization server records the disable. The receiver meets its bound if that gap does not exceed the computed delay. It SHOULD also log the maximum age in effect with each verification result, and the security section says receivers SHOULD treat such changes as security relevant. The draft describes this test and reports no results from running it.

    A disabled agent's old tokens still pass every check; a DNS draft now asks receivers to time how longBI-FIND-2026-2345

  2. AAuth's premise is that an OAuth client or an API key has no identity of its own and that SPIFFE and WIMSE stop at one trust domain, so it gives each agent one: an identifier of the form aauth:local@domain issued by an Agent Provider and bound to the agent's own signing key by a short-lived agent token (the draft recommends Ed25519, which every party must support), so that 'every agent has its own cryptographic identity' and 'the first API call to a resource is the registration'. No credential in the protocol is a bearer credential; every request is signed, and a resource can challenge for more at any time. In its lightest mode the draft pitches this as a drop-in for API keys: 'this replaces API keys with cryptographic identity.' What an agent token attests about the software behind the key is left to companion specifications.

    AAuth gives every agent its own key, then keeps the delegation chain out of the token on purposeBI-FIND-2026-0365

  3. Asor's chain names its principal at every hop but cannot show that the principal still holds the grant. Every token's sub must carry the same value, the party "whose grant the chain acts under"; a required client_id names the agent acting at each hop, and step 3 denies a chain whose sub changes. A row added to Table 1 in -03, credited to Cliff Singletary, records the limit: nothing in a chain says whether the principal named in sub still holds the authority the root token granted. A principal who has left, lost the role or had the grant withdrawn leaves every unexpired, unrevoked chain verifying clean, because a status list revokes tokens and nothing in the chain revokes the principal. A deployment that must know checks standing when the root token is issued, which is also where the draft puts approval. Where the gap after minting is unacceptable, an online server in the call path checks again at use. Approval at the root has its own limit: verification shows that no hop exceeded the root, "never that the root was right".

    Asor's agent-chain draft now says a chain keeps verifying after its principal leaves, until expiry or revocationBI-FIND-2026-1335

›Cited in 28 briefings

Every briefing that discusses this term, newest first.

  1. A disabled agent's old tokens still pass every check; a DNS draft now asks receivers to time how long

    Draft · · IETF Datatracker · Sriram Shankarlingam, HexaEight · Oct 20, 2026

  2. AAuth gives every agent its own key, then keeps the delegation chain out of the token on purpose

    Draft · · IETF Datatracker · Dick Hardt, Hellō · Sep 25, 2026

  3. Asor's agent-chain draft now says a chain keeps verifying after its principal leaves, until expiry or revocation

    Draft · · IETF Datatracker · Rafael Asor · Oct 28, 2026

  4. Transaction Tokens carry who started a request through internal services and define no list of the hops in between

    Draft · · IETF OAuth Working Group · Atul Tulshibagwale et al. · Jul 30, 2026

  5. OAuth's draft for crossing trust domains is approved for publication and never says agent

    Draft · · IETF OAuth Working Group · Arndt Schwenkschuster et al. · Jul 19, 2026

  6. Agent-audit draft now admits a call on the principal and the current agent, as RFC 8693 requires

    Draft · · IETF Datatracker · Shameer Thaha et al. · Oct 19, 2026

  7. An agent-evidence draft no longer checks every actor in a delegation chain before a service admits a call

    Draft · · IETF Datatracker · Shameer Thaha et al. · Oct 19, 2026

  8. WIMSE's agent draft adds how an agent's access ends and where policy engines fit, without a new requirement

    Draft · · IETF WIMSE Working Group · Pieter Kasselman et al. · Sep 15, 2026

  9. An agent audit draft says the agent is not trusted, yet lets the agent's operator run the audit store

    Draft · · IETF Datatracker · Mirja Kühlewind et al. · Sep 7, 2026

  10. Sweeney's delegation draft concedes its server is trusted with users' credentials, and drops an interop claim

    Draft · · IETF Datatracker · Kieran Sweeney · Oct 13, 2026

  11. Cross-App Access draft warns that one stolen sign-in token could reach every app the identity provider permits

    Draft · · IETF Datatracker · Aaron Parecki et al. · May 21, 2026

  12. Agent delegation records not anchored in trusted time become deniable once timestamps can be forged

    Draft · · IETF Datatracker · V.K. Uppalapati · Sep 29, 2026

  13. Niyikiza's agent tokens narrow to a tool's arguments offline; revision -02 breaks -01 implementations

    Draft · · IETF Datatracker · Niki Aimable Niyikiza · Jun 15, 2026

  14. OAuth tokens prove an agent's rights, not a person's consent; revision 05 adds a day-long grant with spend caps

    Draft · · IETF Datatracker · Meiling Chen et al. · Oct 27, 2026

  15. A cross-domain transaction-token grant now names AI agents as a use and keeps the call chain at home

    Draft · · IETF Datatracker · George Fletcher et al. · Sep 10, 2026

  16. Six companies' agent-auth draft, since adopted by the WIMSE group, keeps credentials out of the model's reach

    Draft · · IETF Internet-Draft (individual submission, since adopted by the WIMSE working group) · Pieter Kasselman et al. · Jul 6, 2026

  17. An OAuth agent profile has the token server block any widening of scope, and leaves revoking the chain optional

    Draft · · IETF Datatracker · Sanjeev Kumar, Orchestrum Technologies LLP · Aug 30, 2026

  18. MCP's enterprise login extension says IdP revocation is immediate. Its own spec does not back that up

    Article · · Model Context Protocol (modelcontextprotocol.io) · MCP project et al. · Jun 18, 2026

  19. OAuth handles an agent in one company, working while you watch. The OpenID paper says that is where it stops.

    Article · · OpenID Foundation · Tobin South et al. · Oct 2025

  20. The IETF's workload-identity architecture gives AI agents one subsection, calling an agent a delegated workload

    Draft · · IETF Datatracker · Joseph Salowey et al. · Jul 6, 2026

  21. A July OAuth draft keeps agent identity stable as keys rotate

    Draft · · IETF Datatracker · Karl McGuinness · Jul 5, 2026

  22. Batch delegation assigns each permission to an agent, then filters at token exchange

    Draft · · IETF Datatracker · Yuan Ni et al. · Jul 3, 2026

  23. One implementer found a transaction-token rule that let any scope through; the OAuth group fixed it in weeks

    Draft · · IETF Datatracker · Dhruv Agnihotri · Jun 23, 2026

  24. A June OAuth draft logs each hand-off between AI agents, but the per-hop limits the act claim lacks are optional

    Draft · · IETF Datatracker · Dapeng Liu et al. · Jun 8, 2026

  25. Atlassian's first draft for offline delegation banned bearer handles; revision 05 just recommends sender binding

    Draft · · IETF Datatracker · Larry Zhu et al. · May 22, 2026

  26. A cross-domain agent trust draft requires the call chain, then leaves its format to a version that never came

    Draft · · IETF Datatracker · Daniel King et al. · Mar 15, 2026

  27. A Huawei agent draft keeps one 'master agent' as gatekeeper and drops the line that had servers re-check it

    Draft · · IETF Datatracker · Yuan Ni et al. · Feb 28, 2026

  28. The July 2025 Cross App Access draft lets the IdP decide app-to-app access and treats agents as plain clients

    Draft · · IETF Datatracker · Aaron Parecki et al. · Jul 2, 2025

In our columns

  1. Reused OAuth is winning the agent standards race and leaves sub-agent narrowing to vendors

    Column · Sep 30, 2026

Related terms

What appears alongside OAuth Identity and Authorization Chaining Across Domains in the same briefings.

Transaction TokensOAuth Token ExchangeBatch Authorization DelegationOAuthAgent Identity Management SystemJWT Authorization GrantJWT Bearer GrantIdentity Assertion Authorization Grant

Was this explainer useful?

Share this explainer

Pass it on. Every claim has a permanent number, so what you share today still points at the same words next year.