Skip to content
IAM Insights

Explainers · Glossary

DPoP

RFC 9449 · Demonstrating Proof of Possession · sender-constrained tokens

By Boston Identity · Explained · 7 min read

The short answer

DPoP (Demonstrating Proof of Possession) is an internet standard, RFC 9449, that ties an OAuth access token to a key the client holds. Each request carries a fresh proof signed with that key, so a copied token fails without it.

What it is
A protocol
Defined by
IETF, RFC 9449 (Standards Track, 2023)
Who it's for
Teams running OAuth APIs, browser apps and agents
In our briefings
87By month of the source: 2025-12: 1, 2026-01: 1, 2026-02: 6, 2026-03: 7, 2026-04: 5, 2026-05: 4, 2026-06: 8, 2026-07: 9, 2026-08: 2, 2026-09: 11, 2026-10: 13, 2026-11: 11by month, last 12
01

What is DPoP?

DPoP is a published standard from the Internet Engineering Task Force (IETF): RFC 9449, issued in September 2023 on the Standards Track.S1 It extends OAuth, the framework apps use to get access tokens for APIs. Its job is to sender-constrain access and refresh tokens, which means tying each token to a key pair so that only the party holding the private key can use it.§1

The problem it answers is the ordinary bearer token, which any party in possession of it can use.§2 Tokens have leaked through bugs in other layers of the software stack and through attacks on OAuth implementations themselves. The RFC calls DPoP "a general defense in depth against the impact of unanticipated token leakage".§2 It works at the application level, in an HTTP header, so it fits clients where mutual TLS is not available, such as a single-page app running in a browser.§1 DPoP is not a way to authenticate the client, and it must always be used with HTTPS.§2, §3

Not client authentication, which proves which app is calling. A DPoP proof shows only that the caller holds a key, and the RFC says a valid proof alone is not enough to make an access decision.§4

02

How does DPoP work?

Three parties take part: the client (the app or agent that wants data), the authorization server that issues tokens, and the resource server, the API that holds the data. The client signs with a private key whose public half it chose itself.§4.2

Figure steps
  1. The client signs a proof for its token request
  2. The authorization server binds the token to the key
  3. The client sends the token with a new proof
  4. The resource server checks the proof and the binding
  5. The resource server serves or refuses
Fig. 1

The client proves its key once to get a bound token from the authorization server, then proves it again, with a fresh proof tied to that token, on every call to the resource server.

Drawn from RFC 9449, section 3 and section 7.1

Swipe for the full figure →

  1. The client signs a proof for its token request

    When the client asks the authorization server for an access token, it attaches a DPoP proof in a DPoP header.§3 The proof is a short JWT (JSON Web Token) signed with the client's private key, with the public key in its header.§4.2 Its payload must hold a unique identifier (jti), the HTTP method (htm), the target URI (htu) and a creation time (iat).§4.2 This applies to every grant type, including a refresh.§5

  2. The authorization server binds the token to the key

    The authorization server checks the proof and ties the access token it issues to the public key in that proof.§5 It must signal this to the client by returning a token_type of DPoP.§5 For a JWT access token, the binding is a hash of the public key, the jkt value inside the cnf claim; a resource server can also learn it through token introspection.§6.1, §6.2 A refresh token issued to a public client, one with no credentials of its own, must be bound to the same key.§5

  3. The client sends the token with a new proof

    To call the API, the client sends the access token in the Authorization header under the DPoP scheme, plus a new proof in the DPoP header.§7.1 Each HTTP request requires a unique proof.§4 This proof must also carry ath, a hash of the access token, which ties the proof to that one token.§4.2, §7

  4. The resource server checks the proof and the binding

    The resource server must check the proof's signature, its type and its algorithm, and that htm and htu match the request it received.§4.3 It must check that the proof was created within an acceptable window and that ath matches the token presented.§4.3 It must also check that the public key in the proof is the key the token is bound to.§7.1

  5. The resource server serves or refuses

    The resource server must not grant access unless all of these checks succeed.§7.1 It can refuse with a 401 response carrying a DPoP challenge, for example one that reports an invalid key binding.§7.1 The access token itself must still be valid in every other respect.§3

03

Why does DPoP matter now?

The rule books now lean toward key-bound tokens. RFC 9700, OAuth's security best current practice, says authorization and resource servers should sender-constrain access tokens, naming mutual TLS and DPoP as mechanisms.1 Revision -15 of the OAuth 2.1 draft is firmer about refresh tokens for public clients: the authorization server must catch their replay, either by binding them to the client instance or by rotating them, with DPoP and mutual TLS given as the binding examples.2 NIST's final token guidance, published with CISA on 15 September 2026, recommends binding tokens to the client's key only for workloads, and all of that guidance is voluntary.3

Agent tooling is starting to carry it. Version 2.1.0 of the official MCP TypeScript client, released 23 September, can bind access tokens with DPoP, but only when the client opts in, and the MCP proposal behind it was still an open pull request on 9 October.4 Hartman's credential-broker draft recommends, as its default, that an agent hold a short-lived token bound to its own key with DPoP.5 In an adoption call that closed on 5 October, two of DPoP's co-authors objected to a second key-binding draft in the OAuth working group, saying a second sanctioned mechanism would hurt interoperability.6

Support is not the same as enforcement. Auth0's Go JWT middleware v3 ships in a default mode that still accepts plain bearer tokens beside DPoP ones.7 In the v3.0.0 code we checked, a proof is accepted for 300 seconds by default and the library keeps no record of proof ids it has already seen.8

04

Where does it fit?

Compared withHow DPoP differs
Bearer tokenA bearer token can be used by any party that holds it.§2 A DPoP-bound token also needs a proof signed with the client's private key on each request.§7.1
mTLSCertificate-bound tokens over mutual TLS (RFC 8705) bind at the transport layer, which the RFC says can give strong message integrity, authenticity and replay protection.§11 DPoP binds at the application layer, which suits browser apps where TLS client authentication gives a poor user experience.§1
Sender-constrainingSender-constraining is the goal of tying a token to the client it was issued to. DPoP is one way to reach it at the application level,§1 and RFC 9700 names mutual TLS as another.1
WIMSE Workload Proof TokenRevision -07 of the WIMSE draft describes the Workload Proof Token as inspired by DPoP, binding the target, the workload identity token and any included authorization-context tokens.9 That revision leaves open whether to use it or an HTTP Message Signatures profile, while DPoP is a published RFC for OAuth tokens.9,S1
05

What goes wrong?

  • Hostile code inside the client defeats it. If an attacker can run code in the client's context, for example through cross-site scripting (XSS), the RFC says DPoP's security is no longer guaranteed.§11.4 A key that cannot be exported stops the attacker from carrying it away, but the attacker can still create new proofs while the client is online.§11.4 Such code can also start a silent authorization grant in an iframe and tie the resulting tokens to the attacker's own key.§11.4
  • A captured proof can be replayed at the same endpoint and method, so servers must accept proofs only for a limited time after creation.§11.1 Servers can also store each proof's jti to reject reuse, but the RFC says that may not be feasible when several servers sit behind one endpoint with no shared state.§11.1
  • Someone in control of the client, including its legitimate user, can sign proofs with future timestamps and copy them to another machine.§11.2 Unpredictable nonces from the server can prevent this, and deployments without them should not issue long-lived DPoP access tokens.§11.2
  • A proof covers only the HTTP method and URI, not the body or general headers, and relies on TLS for message integrity.§11.7
  • A resource server that supports both schemes must reject a DPoP-bound token sent as a bearer token, but a server unaware of DPoP would most presumably accept it.§7.2 An authorization server may also return a plain bearer token, and the client must discard it if that protection is important to the application.§5
06

What should a security leader ask?

  1. 01

    Which of our APIs still accept a DPoP-bound token sent as a plain bearer token, and when do we switch them to reject it?

  2. 02

    How long does each of our resource servers accept a DPoP proof after it was created, and does it reject a proof id it has already seen?

  3. 03

    Where do our browser apps, mobile apps and agents keep their DPoP private keys, and can code running in the app export them?

  4. 04

    Do our authorization servers send nonces, and if not, how long do our DPoP access tokens live?

  5. 05

    Which agent frameworks and SDKs (software development kits) we use have DPoP turned on, not just available?

07

Sources

›Evidence: 9 claims from 8 briefings
  1. 1BI-FIND-2025-0352 · RFC 9700 bans OAuth's password grant and loose redirect matching; binding access tokens stays a SHOULDFor access tokens the RFC stops at SHOULD. Authorization and resource servers should sender-constrain them, tying each token to a key the client holds with a mechanism such as mutual TLS (RFC 8705) or DPoP (RFC 9449), and should restrict each token to one resource server or a small set. The threat is an attacker who gets hold of a token, for example by running a counterfeit resource server or compromising a real one. The authors concede that "architecture and performance reasons" may rule these measures out in some deployments, and that a bound token is undermined once the attacker also has the key. They also decline to rely on clients to spot a counterfeit resource server, because many clients fail to implement checks such as state: relying on them "is likely to fail as well".
  2. 2BI-FIND-2026-1109 · OAuth 2.1's March draft builds in the security best practice, but 2.1-only clients can fail with older serversRevision -15 of OAuth 2.1 turns the advice in OAuth 2.0's security best current practice (RFC 9700) into the base protocol. The authorization code grant now carries PKCE: clients are REQUIRED to send a code challenge and authorization servers MUST enforce it, unless the client is confidential and the server has reasonable assurance that it implements the OpenID Connect nonce properly. Redirect URIs are registered in full and must match exactly, except for the port on loopback redirects. The implicit and password grants are left out, clients MUST NOT send access tokens in a URI query parameter, and for public clients the authorization server MUST catch refresh-token replay, either by binding the token to the client instance (DPoP and mutual TLS are the examples given) or by rotating it on every refresh.
  3. 3BI-FIND-2026-0700 · NIST and CISA advise one-hour tokens, 90-day keys and hardware key storage; AI agents get one paragraphNIST's final guidance on protecting identity tokens, published with CISA on 15 September 2026, turns token hygiene into numbers. Access and identity tokens should be valid for no more than an hour; signing keys for high-impact systems should be active no more than 90 days and others less than a year; and from moderate impact up, signing keys must sit in hardware-backed or otherwise isolated storage, which a cloud key service satisfies. Keys that sign federal tokens must not sign anything else, workload identities must use tightly scoped short-lived tokens, and tokens must never be written to logs. Binding tokens to the client's key is recommended only for workloads, and the report concedes that stateless tokens cannot always be revoked before they expire. All of it is voluntary.
  4. 4BI-FIND-2026-0684 · The official MCP TypeScript SDK now binds tokens to keys, ahead of the spec it implementsVersion 2.1.0 of the official MCP TypeScript client, released 23 September, can bind an access token to a key the client holds (DPoP, RFC 9449), so a token copied out of a log or a proxy is useless without that key. It signs a proof into every token and resource request, handles server nonces, and re-authorizes when a refresh token is bound to a key it no longer has. It is opt-in: nothing happens unless the client implements the new hook, and a token the server issued as a plain bearer token is still sent as one. The MCP proposal it implements, SEP-1932, was still an open pull request on 9 October, so the spec does not yet ask anyone to use it.
  5. 5BI-FIND-2026-0534 · A SANS instructor's broker vaults agents' real keys, but its third delivery model breaks the abstract's promiseHartman's broker splits the decision from the secret. A policy decision point that "MUST never touch credential material" approves, denies or escalates a signed task request naming the target, action, scope and lifetime; a separate delivery point pulls the real credential from a hardware-backed vault, mints something narrower and wipes its memory. Three delivery models follow. In the first, a proxy injects the credential and the agent never sees it. In the second, recommended as the default, the agent holds a short-lived derived token bound to its own key with DPoP. In the third, marked NOT RECOMMENDED, the agent receives the real long-lived credential and it is revoked after the task. So the abstract's promise that agents "never hold real long-lived credentials" holds only if the third model is never used, and two sections disagree on which model is the fallback for services that cannot issue short-lived tokens.
  6. 6BI-FIND-2026-0862 · The OAuth group split over a second way to bind tokens to keys, and the draft never says what DPoP gets wrongRicher's draft adds a second way to bind an OAuth access token to a client's key, and the working group split over it without the draft stating why. The text says only that it "complements and builds on experience with" DPoP and mutual TLS; the guidance on when to use which is an editor's note, the security and privacy sections are marked to-do, there are no test vectors, and no implementation is named. Nine people supported adoption in the call that closed on 5 October, five of them the draft's own authors. Four objected: Dick Hardt said the group had already solved key binding with DPoP, and Michael Jones and Brian Campbell, two of DPoP's co-authors, said a second sanctioned mechanism would hurt interoperability. No decision had been posted a week later.
  7. 7BI-FIND-2026-1089 · Auth0's Go JWT middleware v3 adds DPoP but accepts a reused proof for five minutes, and bearer tokens by defaultv3 ships with DPoP in its permissive mode. DPoPAllowed, the default, accepts plain bearer tokens alongside DPoP ones, while DPoPRequired, which rejects bearer tokens outright, is what Dawar calls "the ultimate end goal for maximum API security." In his section on what v2 lacked, he describes the risk DPoP answers: "if someone steals the token, they effectively become you." An API that moves to v3 and leaves that setting alone still accepts a plain bearer token from whoever holds it, until its owners switch modes.
  8. 8BI-FIND-2026-1088 · Auth0's Go JWT middleware v3 adds DPoP but accepts a reused proof for five minutes, and bearer tokens by defaultDawar writes that go-jwt-middleware v3 checks a DPoP proof's signature, its HTTP method and URL, the access-token hash and the key thumbprint, and that it "checks the proof's age to prevent replay attacks." In the v3.0.0 code we checked, that age check is the whole replay defence: by default a proof is accepted for 300 seconds after it was made, and the library keeps no record of proof ids (jti) it has already seen, although the repository's migration guide lists "Replay protection via jti and iat claims". RFC 9449 requires the time limit and leaves a single-use check to the server's choice. We read this to mean that anyone who copies a token and its proof from a log or a proxy can send the same request to the same method and URL within those five minutes, and it passes every check the post lists.
  9. 9BI-FIND-2025-0493 · WIMSE leaves its application proof choice open, with different integrity coverageRevision -07 leaves the application mechanism open between a Workload Proof Token (WPT), inspired by OAuth Demonstrating Proof of Possession (DPoP), and an HTTP Message Signatures profile. The WPT binds the target, the Workload Identity Token and any included authorization-context tokens. The signatures profile requires the HTTP method and request target to be signed, plus specified headers when present, and requires a Content-Digest for a message body. Responses may be signed; this is optional. The draft warns that the DPoP-inspired alternative does not protect major portions of requests and responses against an actively malicious middlebox. Both application options still require a secure channel.

›In our briefings

The claim each of the newest briefings on DPoP would have you leave with.

  1. Okta's post says version 1.1 of the Okta Identity as a Service (IDaaS) Security Technical Implementation Guide (STIG), which Okta first published with the U.S. Defense Information Systems Agency (DISA) in March 2025, adds five checks. Two are about API tokens, which the authors call "almost always privileged and sensitive." Each token must be configured with Network Zones to restrict it to known networks, which the post says helps confirm that a compromised token "cannot be used from an unapproved IP", and each must be created under a dedicated user account that is not tied to an administrator. The other three require IP allow or deny rules in the Global Session policy, in line with the Access Control policy for Okta, network zones that block anonymized proxies according to the organization's policy, and network zones in each application's authentication policy, where the default is to allow internet access. The authors say the checks are specific to Okta but "the same approach should be used for other service offerings." They also say these are the kinds of checks Okta used to help prevent attacks such as the Salesloft Drift and Gainsight attacks, and the post gives no detail on how beyond links to two other Okta posts.

    Okta's updated STIG says API tokens must be limited to known networks and kept off admin accountsBI-FIND-2026-2390

  2. 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

  3. 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

›Cited in 87 briefings

Every briefing that discusses this term, newest first.

  1. Okta's updated STIG says API tokens must be limited to known networks and kept off admin accounts

    Article · · Okta Security · Rob Gil et al. · Feb 3, 2026

  2. 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

  3. 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

  4. 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

  5. The OAuth working group split over a token-binding draft that never says what DPoP lacks

    Draft · · IETF Datatracker · Justin Richer et al. · Jul 28, 2026

  6. 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

  7. Sello has the services an AI agent calls sign its receipts; the agent can't fake them, but it can go around them

    Paper · · arXiv · Juan Figuera · Jun 2, 2026

  8. 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

  9. Okta and Google tie an agent's calls to the user, while GitHub's shared Copilot automations now act as Copilot

    Release tracker · · 11 vendors' release notes · Oct 2026

  10. An OAuth client gets to name who vouches for it, and removing one waits on server caches

    Draft · · IETF Datatracker · Karl McGuinness · Sep 28, 2026

  11. An OAuth draft lets the API, not the agent, describe the approval, and now name the key the token is bound to

    Draft · · IETF Datatracker · Yaroslav Rosomakho et al. · Jun 25, 2026

  12. 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

  13. Cisco's SD-Card rewrite pins MCP tools to what a registry signed; its security sections still read TODO

    Draft · · IETF Datatracker · Suhas Nandakumar et al. · Nov 1, 2026

  14. 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

  15. OAuth's draft for token requests that wait on a person now gives the client a link for the user to approve

    Draft · · IETF Datatracker · Guilherme de Oliveira Niero et al. · Sep 23, 2026

  16. 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

  17. 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

  18. Lee's draft would hold an agent's action at the boundary until a receipt for that exact action passes every check

    Draft · · IETF Datatracker · Yong Bok Lee, Meridian Verity Group · Jun 5, 2026

  19. Agent URI draft -04 ties an agent's name to a key document and bars OAuth servers from trusting the name alone

    Draft · · IETF Datatracker · Yaswanth Narvaneni et al. · Oct 20, 2026

  20. 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

  21. One global registry for AI agent identities, and no one allowed to switch an identity off

    Draft · · IETF Datatracker · Christopher Drake, 1id.com · Oct 27, 2026

  22. 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

  23. 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

  24. The IETF's stronger workload proof is built for the proxy in the middle, and it still leaves replay protection optional

    Draft · · IETF Datatracker · Joseph A. Salowey et al. · Oct 20, 2026

  25. ChainIT's design has the payment rail rebuild each payment's hash before money moves; the paper reports no test

    Paper · · arXiv · Jeremy Blackburn et al. · Oct 16, 2026

  26. 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

  27. An OAuth profile for AI agents makes every hand-off cut privileges, but a shorter expiry alone counts as a cut

    Draft · · IETF Datatracker · Angel Cruz · Feb 7, 2026

  28. An attested OAuth client is one somebody vouched for; revision 12 says to trust its claims only as far as the voucher

    Draft · · IETF Datatracker · Tobias Looker et al. · Oct 13, 2026

  29. In Huawei's reworked OAuth draft, evidence signed by the token's subject can narrow access but never widen it

    Draft · · IETF Datatracker · Cheng-Kang Chu et al. · Sep 1, 2026

  30. Huawei's OAuth draft for teams of agents enforces its firm limit on authority only in an optional mode

    Draft · · IETF Datatracker · Yurong Song et al. · Jul 1, 2026

  31. OpenA2A's draft aims to keep secrets from the agent, and says no implementation yet supports its new controls

    Draft · · IETF Datatracker · Abdel Fane, OpenA2A · Oct 2, 2026

  32. The official MCP TypeScript SDK now binds tokens to keys, ahead of the spec it implements

    Article · · WorkOS Blog · Maria Paktiti · Sep 29, 2026

  33. TPX replaces an app's embedded AI provider key with a capped, metered grant the user approves at their provider

    Article · · tokenpony.dev · Alex Polvi · Jul 20, 2026

  34. OWASP's agent Top 10 blames identity systems built for people, and does not say how it chose or ranked its risks

    Article · · OWASP GenAI Security Project · John Sotiropoulos et al. · Dec 9, 2025

  35. NIST and CISA advise one-hour tokens, 90-day keys and hardware key storage; AI agents get one paragraph

    Article · · NIST (CSRC) · Ryan Galluzzo et al. · Sep 15, 2026

  36. 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

  37. A survey rates only workload identity ready for agents, and argues no standard can tell what an agent intends

    Paper · · arXiv · Takumi Otsuka et al. · Apr 25, 2026

  38. Coetzee's agent token is good for one declared call, and the hard part is hashing that call the same way twice

    Draft · · IETF Datatracker · Rudolf Jacobus Coetzee · Jul 19, 2026

  39. A SANS instructor's broker vaults agents' real keys, but its third delivery model breaks the abstract's promise

    Draft · · IETF Datatracker · Kenneth G. Hartman · Mar 29, 2026

  40. Transaction tokens are nearly a standard, and there is still no agreed place in them for the agent

    Podcast · · Identity at the Center · #449 - Decoded - Transaction Tokens with George Fletcher · Sep 21, 2026

  41. A draft token hides some claims, works with old servers unchanged, and leaves agent delegation claims visible

    Draft · · IETF Datatracker · Su-hyeon Forten Lee · Oct 3, 2026

  42. Cisco's SD-Card puts an A2A Agent Card in a selective-disclosure token, and leaves its registry out of scope

    Draft · · IETF Datatracker · Suhas Nandakumar et al. · Mar 2, 2026

  43. The IETF's simplest workload proof proves the key and leaves the request unprotected, and the draft says so itself

    Draft · · IETF Datatracker · Brian Campbell et al. · Aug 27, 2026

  44. 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

  45. The 'intent' in the Agentic JWT draft is a workflow step, and the agent reports its own fingerprint

    Draft · · IETF Datatracker · Abhishek Goswami · Jul 1, 2026

  46. Highflame's open-source ZeroID narrows an agent's token at every hop, a step we did not find in Uber's write-up

    Article · · Highflame (GitHub) · Sep 25, 2026

  47. IPC passes a delegation chain between services; its signature proves only the last signer

    Draft · · IETF Datatracker · Ravi Kant Sharma, Ericsson · Jul 23, 2026

  48. The browser OAuth draft favors a backend that keeps tokens out of JavaScript

    Draft · · IETF Datatracker · Aaron Parecki et al. · Jul 6, 2026

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

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

  50. 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

  51. An OAuth draft lets device checks set an app's scopes; a first review found a copied app could reuse the result

    Draft · · IETF Datatracker · Efe Kahraman, Mekarge · Jun 9, 2026

  52. 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

  53. 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

  54. Protocol 427 offers agents a reusable spending allowance, then ties each signed proof to one request

    Draft · · IETF Datatracker · John Paul McGraw et al. · May 6, 2026

  55. Three vendors put MCP controls in three different places in May, and only Auth0's was GA

    Release tracker · · 10 vendors' release notes · May 2026

  56. An Imperial and Oxford protocol would give AI agents one approved use of a secret, never the secret itself

    Paper · · arXiv · Xiaohang Yu et al. · Apr 27, 2026

  57. A 'decentralized' agent identity draft blocks risky calls when its Registry is down, and did:aip is taken

    Draft · · IETF Datatracker · Paras Singla · Apr 17, 2026

  58. An agent certificate draft ties access to a trust score set by the issuing CA, usually the agent operator's own

    Draft · · IETF Datatracker · Raza Sharif, CyberSecAI · Apr 10, 2026

  59. Microsoft and Google made AI agent identity generally available in April; Okta's notes put its GA in May

    Release tracker · · 8 vendors' release notes · Apr 2026

  60. OpenID agent-claims draft adds a provider-set user, application or agent label; scores stay each provider's own

    Draft · · IETF Datatracker · Raza Sharif, CyberSecAI · Mar 26, 2026

  61. Checking each agent call against the user's own task stopped all 634 off-task calls the authors wrote

    Paper · · arXiv · Reshabh K Sharma et al. · Mar 17, 2026

  62. 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

  63. An MCP signing draft shows a client who wrote a tool, and admits a poisoned tool can still be validly signed

    Draft · · IETF Datatracker · Raza Sharif, CyberSecAI · Mar 14, 2026

  64. OAuth 2.1's October draft bans plain PKCE and adds browser-app rules; 2.1-only clients can fail on older servers

    Draft · · IETF Datatracker · Dick Hardt et al. · Mar 2, 2026

  65. OAuth's in-app login draft lets a native app collect the password itself, and leaves each server to vet the app

    Draft · · IETF Datatracker · Aaron Parecki et al. · Feb 28, 2026

  66. Auth0's Go JWT middleware v3 adds DPoP but accepts a reused proof for five minutes, and bearer tokens by default

    Article · · Auth0 Blog · Kunal Dawar · Feb 27, 2026

  67. A draft binds workload credentials to their host via a local agent, and leaves that agent's compromise undecided

    Draft · · IETF Datatracker · Ram Krishnan et al. · Feb 26, 2026

  68. OpenPort holds agent writes as drafts for an operator by default; its own tests check less than its spec asks

    Paper · · arXiv · Genliang Zhu et al. · Feb 22, 2026

  69. OAuth's cross-device guidance ranks FIDO first, then CIBA backchannel login, and the device-code flow last

    Draft · · IETF Datatracker · Pieter Kasselman et al. · Jan 23, 2026

  70. Huawei-led OAuth draft lets clients chain narrower tokens offline but leaves revocation delivery out of scope

    Draft · · IETF Datatracker · Li Ruochen et al. · Oct 20, 2025

  71. Rosenberg and White's AAuth now has the resource server, not the agent, write and sign the approval text

    Draft · · IETF Datatracker · Jonathan Rosenberg et al. · Oct 20, 2025

  72. WIMSE leaves its application proof choice open, with different integrity coverage

    Draft · · IETF Datatracker · Brian Campbell et al. · Oct 16, 2025

  73. A refresh-token thief who waits for the app to close slips past reuse detection

    Article · · Auth0 Blog · Maria Vasilevskaya, Okta · Oct 8, 2025

  74. Agentic JWT binds an agent's token to its task and its own code, but it is one author's draft, not a standard

    Paper · · arXiv (cs.CR preprint 2509.13597) · Abhishek Goswami · Sep 16, 2025

  75. OpenID's credential issuance standard is OAuth for wallets, and assumes a trusted wallet is an app on a phone

    Article · · OpenID Foundation · Torsten Lodderstedt et al. · Sep 16, 2025

  76. Auth0's DPoP explainer leaves out the access-token hash that RFC 9449 requires on every API call

    Article · · Auth0 Blog · Andrea Chiarelli · Aug 25, 2025

  77. An early MCP security checklist was never tested, and the protocol has since adopted only part of it

    Paper · · arXiv · Vineeth Sai Narajala et al. · Apr 11, 2025

  78. A rogue OAuth server can collect a client's signed assertion and replay it at an honest server the client also uses

    Article · · IACR Cryptology ePrint Archive · Pedram Hosseyni et al. · Apr 7, 2025

  79. RFC 9728 tells clients where an API's tokens come from and leaves most of the trust decision to them

    Article · · RFC Editor · Michael B. Jones et al. · Apr 2025

  80. FAPI 2.0 issues only sender-constrained access tokens and leaves public clients out

    Article · · OpenID Foundation · Daniel Fett et al. · Feb 22, 2025

  81. RFC 9700 bans OAuth's password grant and loose redirect matching; binding access tokens stays a SHOULD

    Article · · IETF (RFC Editor) · Torsten Lodderstedt et al. · Jan 2025

  82. In May 2023, Parecki described OAuth 2.1 as consolidation and DPoP as token protection

    Video · · Identiverse - A CRA Resource · Jun 17, 2024

  83. RFC 9470 lets an API demand a fresher or stronger sign-in, with no shared scale for 'stronger'

    Article · · RFC Editor · Vittorio Bertocci et al. · Sep 8, 2023

  84. DPoP token binding depends on checks at the resource server

    Article · · RFC Editor · Daniel Fett et al. · Sep 2023

  85. CIBA's human-approval step carries a short matching code, not a description of the action

    Article · · OpenID Foundation · Gonzalo Fernandez Rodriguez et al. · Sep 1, 2021

  86. RFC 6750 tokens work for whoever holds them; RFC 9700 now urges key binding and bars them from query strings

    Article · · RFC Editor · Michael B. Jones et al. · Oct 1, 2012

  87. RFC 6749 admits a server can't tell who presents a token; the fixes arrived in 2020 and 2023

    Article · · RFC Editor · D. Hardt et al. · Oct 2012

In our columns

  1. Make stolen tokens useless before trying to tell if an AI is using them

    Column · Oct 29, 2026

Related terms

What appears alongside DPoP in the same briefings.

WIMSE Workload Proof TokenAAuth (Hardt)OAuth 2.0 for Browser-Based ApplicationsBearer tokenOAuth 2.0 AI Agent Instance ProfileSender-constrainingRich Authorization RequestsOAuth Token Exchange

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.