Skip to content
IAM Insights

Explainers · Glossary

GNAP

Grant Negotiation and Authorization Protocol

An IETF authorization protocol designed to replace OAuth 2.0's grant model with a single, extensible negotiation between client and authorization server, folding in capabilities (like PAR and device flow) that OAuth bolted on as separate extensions.

reference →

In our briefings

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

  1. 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

  2. ChainIT's protocol ties each payment's policy decision, budget reservation and single-use execution credential to one canonical transaction digest: a hash over every material field of the instruction that will actually be transmitted. Inside the band an agent's grant allows (in the paper's invented example, approved cloud-services vendors up to $5,000 per transaction and $25,000 a month), the deterministic control plane decides with no human involved. Past the step-up boundary, such as a new payee, a changed destination or a higher value, a person with independent authority must approve that exact digest; in the authors' words, "the approver approves a digest, not a screen." Either way, the adapter at the payment rail "must reconstruct the canonical transaction digest from the rail instruction" and check it against the policy decision, the reservation and any approvals before it consumes the credential, before or atomically with submission. A beneficiary account changed after approval gives a different digest, so the existing approvals no longer apply. The authors describe this as the EU's dynamic-linking rule for strong customer authentication extended from amount and payee to every field, and say it shares the goals of transaction binding and single use with Lan et al.'s addition to Google's AP2.

    ChainIT's design has the payment rail rebuild each payment's hash before money moves; the paper reports no testBI-FIND-2026-1374

  3. In Bu's review method, each security claim an agent protocol makes, such as who authorized the task, which running instance is acting, what scope was delegated, whether a human approved and what evidence the action left, should get its own row naming what carries it, who verifies it, what it is bound to, how fresh it must be and what happens when it fails. Claims cannot be combined into a stronger one than their parts, and "a receipt, log entry, or audit record proves only the statement it records". A protocol that cannot fill a row can instead mark the claim as inherited, out of scope or left to deployment policy.

    Bu's checklist names a verifier for each agent-protocol claim and checks 'the same agent as before' separatelyBI-FIND-2026-0860

›Cited in 15 briefings

Every briefing that discusses this term, newest first.

  1. 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

  2. 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

  3. Bu's checklist names a verifier for each agent-protocol claim and checks 'the same agent as before' separately

    Draft · · IETF Datatracker · Songbo Bu · Oct 14, 2026

  4. Schrock's receipts prove a named key approved one exact agent action, and the draft lists at length what they cannot prove

    Draft · · IETF Datatracker · Iman Schrock, EMILIA Protocol · Sep 12, 2026

  5. 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

  6. A draft ties what each AI agent may read to what it may do, and still can't take a grant back

    Draft · · IETF Datatracker · Hamdy Atakora · Sep 21, 2026

  7. 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

  8. Mission-Bound OAuth's author now says the mission object does not belong in OAuth

    Article · · Control Plane by Karl McGuinness · Karl McGuinness · Mar 15, 2026

  9. xChk ships human OIDC portfolios and a separate agent approval API

    Paper · · arXiv · Sean MacGuire · Jul 15, 2026

  10. 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

  11. 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

  12. 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

  13. 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

  14. An MIT-led paper has the user sign the agent's delegation token and lets OpenID Connect vouch for both parties

    Paper · · arXiv · Tobin South et al. · Jan 16, 2025

  15. In GNAP (RFC 9635), a client is known by its key, and tokens stay key-bound unless the server flags them bearer

    Article · · RFC Editor · Justin Richer et al. · Oct 1, 2024

Related terms

What appears alongside GNAP in the same briefings.

Rich Authorization RequestsAAuth (Hardt)MacaroonsOpenID Connect for AgentsOASNT-CAIDAgentic JWTNIST SP 800-63XACML