Internet-Draft DAAP February 2026
Kumar Expires 31 August 2026 [Page]
Workgroup:
Web Authorization Protocol
Internet-Draft:
draft-mishra-oauth-agent-grants-00
Published:
Intended Status:
Informational
Expires:
Author:
S. Kumar
Grantex

Delegated Agent Authorization Protocol (DAAP)

Abstract

Artificial intelligence (AI) agents increasingly take autonomous actions — submitting forms, initiating payments, and sending communications — on behalf of human users across third-party services. This document defines the Delegated Agent Authorization Protocol (DAAP), an open, model-neutral, framework-agnostic protocol that specifies: cryptographic agent identity using Decentralized Identifiers (DIDs); a human-consent-based grant authorization flow modelled on OAuth 2.0; a signed JSON Web Token (JWT) grant token format with agent-specific claims; a revocation model with online verification; a hash-chained append-only audit trail; a policy engine for automated authorization decisions; and a multi-agent delegation model with cascade revocation. DAAP fills a gap unaddressed by existing OAuth 2.0 extensions: verifying that a specific human authorized a specific AI agent to perform a specific action, revoking that authorization in real time, and producing a tamper-evident record of what the agent did.

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."

This Internet-Draft will expire on 31 August 2026.

Table of Contents

1. Introduction

Deployed AI agents operate across arbitrary third-party services using credentials and permissions that belong to the human users they serve. Today, no interoperable standard exists for:

  1. Verifying that an agent is who it claims to be

  2. Confirming that a specific human authorized a specific agent to perform a specific action

  3. Revoking that authorization in real time across all active tokens

  4. Producing a tamper-evident record of agent activity

OAuth 2.0 [RFC6749] and its extensions address authorization for applications acting on behalf of users, but were not designed for the AI agent use case, which introduces distinct requirements:

DAAP addresses these requirements as a layered extension to OAuth 2.0 concepts, reusing RFC-standard JWT and JWK primitives wherever possible.

1.1. Requirements Language

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.

1.2. Terminology

The following terms are used throughout this document:

Agent:

An AI-powered software process that takes autonomous actions on behalf of a Principal. An Agent has a persistent cryptographic identity (DID) and must obtain an explicit grant from its Principal before acting on their behalf.

Principal:

The human user who authorizes an Agent to act on their behalf. The Principal is the subject (sub) of any Grant Token issued by the authorization server.

Developer:

The organization or individual who built and operates the Agent. The Developer authenticates to the authorization server using an API key.

Authorization Server:

A server implementing this specification that issues Grant Tokens, maintains the grant registry, and provides the JWKS endpoint for offline verification.

Service:

Any API or platform that receives requests from an Agent. Services MUST verify Grant Tokens before acting on agent requests.

Grant:

A persistent record of permission given by a Principal to an Agent for a specific set of Scopes. A Grant is represented to the Agent as a Grant Token.

Grant Token:

A signed JWT [RFC7519] representing a valid, non-revoked Grant. Grant Tokens are short-lived credentials carrying agent-specific claims defined in Section 5.

Scope:

A named permission string following the format resource:action[:constraint] as defined in Section 3.

DID:

A Decentralized Identifier [DID-CORE] — the Agent's cryptographic identity. In DAAP, Agent DIDs take the form did:grantex:<agent_id>.

Policy:

A rule evaluated by the Policy Engine (Section 10) that automatically approves or denies an authorization request before the consent UI is shown to the Principal.

Anomaly:

A behavioral deviation from an agent's established activity baseline, detected by the runtime monitoring system defined in Section 11.

2. Agent Identity

2.1. DID Format

Every Agent registered with a DAAP-compliant Authorization Server receives a Decentralized Identifier of the form:

did:grantex:<agent_id>

where <agent_id> is a ULID (Universally Unique Lexicographically Sortable Identifier) [ULID] prefixed with ag_.

Example:

did:grantex:ag_01HXYZ123abcDEF456ghi

2.2. Identity Document

The DID resolves to an identity document at the Authorization Server. The document MUST contain the following fields:

{
  "@context": "https://grantex.dev/v1/identity",
  "id": "did:grantex:ag_01HXYZ123abcDEF456ghi",
  "developer": "org_yourcompany",
  "name": "travel-booker",
  "description": "Books flights and hotels on behalf of users",
  "declaredScopes": ["calendar:read", "payments:initiate:max_500"],
  "status": "active",
  "createdAt": "2026-02-01T00:00:00Z",
  "verificationMethod": [{
    "id": "did:grantex:ag_01HXYZ123abcDEF456ghi#key-1",
    "type": "JsonWebKey2020",
    "publicKeyJwk": { "..." : "..." }
  }]
}

2.3. Key Management

Authorization Servers MUST adhere to the following key management requirements:

  • Authorization Servers MUST use RS256 (RSASSA-PKCS1-v1_5 using SHA-256) [RFC7518] for signing Grant Tokens.

  • Private signing keys MUST never be transmitted or stored outside the Authorization Server's trust boundary.

  • Public keys MUST be published at /.well-known/jwks.json as a JWK Set [RFC7517].

  • Key rotation MUST be supported without changing the Agent's DID or invalidating existing, unexpired Grant Tokens. Rotated keys MUST remain in the JWKS until all tokens signed with them have expired.

3. Scope Format and Registry

3.1. Format

Scopes are permission strings of the form:

resource:action[:constraint]

where:

  • resource identifies the data or service being accessed (e.g., calendar, payments, email)

  • action identifies the operation (e.g., read, write, send, initiate, delete)

  • constraint is an optional limiting parameter (e.g., max_500 for a spending limit)

3.2. Standard Scope Registry

The following scopes constitute the normative standard registry. Implementations MUST support all standard scopes that are relevant to the resources they expose:

Table 1
Scope Description
calendar:read Read calendar events
calendar:write Create, modify, and delete calendar events
email:read Read email messages
email:send Send emails on the Principal's behalf
email:delete Delete email messages
files:read Read files and documents
files:write Create and modify files
payments:read View payment history and balances
payments:initiate Initiate payments of any amount
payments:initiate:max_N Initiate payments up to N in the account's base currency
profile:read Read profile and identity information
contacts:read Read address book and contacts

3.3. Custom Scopes

Services MAY define custom scopes using reverse-domain notation per [RFC3986]:

com.stripe.charges:create:max_5000
io.github.issues:create

Custom scopes MUST use reverse-domain notation to avoid collisions with the standard registry.

3.4. Scope Display Requirements

Authorization Servers MUST maintain a human-readable description for each scope in their registry. Consent UIs MUST display human-readable descriptions to Principals, never raw scope strings.

4. Grant Authorization Flow

4.1. Overview

The DAAP grant flow is modelled on the OAuth 2.0 Authorization Code flow [RFC6749] with the following adaptations: the client is always a Developer (identified by an API key), the resource owner is a Principal identified by the Developer's internal user identifier, and the resulting token carries agent-specific claims.

Developer App          Authorization Server       Principal
     |                         |                      |
     | POST /v1/authorize       |                      |
     | {agentId, principalId,  |                      |
     |  scopes, redirectUri}    |                      |
     |------------------------>|                      |
     |                         |                      |
     |<------------------------|                      |
     | {authRequestId,          |                      |
     |  consentUrl}             |                      |
     |                         |                      |
     | redirect user --------------------------------->|
     |                         | consent UI displayed |
     |                         |<---------------------|
     |                         | Principal approves   |
     |                         |                      |
     |<------------------------------------------------|
     | redirectUri?code=AUTH_CODE                      |
     |                         |                      |
     | POST /v1/token           |                      |
     | {code, agentId}          |                      |
     |------------------------>|                      |
     |<------------------------|                      |
     | {grantToken,             |                      |
     |  refreshToken}           |                      |

4.2. Authorization Request

The Developer initiates the flow by sending:

POST /v1/authorize
Authorization: Bearer <api_key>
Content-Type: application/json

{
  "agentId": "ag_01HXYZ123abc",
  "principalId": "user_abc123",
  "scopes": ["calendar:read", "payments:initiate:max_500"],
  "expiresIn": "24h",
  "redirectUri": "https://yourapp.com/auth/callback",
  "state": "<csrf_token>",
  "audience": "https://api.targetservice.com"
}

The audience field is OPTIONAL. When present, it MUST be embedded as the aud claim in the issued Grant Token. The state parameter is REQUIRED and MUST be validated by the Developer's callback handler to prevent CSRF attacks. The redirectUri MUST match a URI pre-registered for the Agent at the Authorization Server.

Authorization Servers MUST reject requests whose redirectUri does not exactly match a pre-registered value for the specified agentId.

Response 200 OK:

{
  "authRequestId": "areq_01HXYZ...",
  "consentUrl": "https://consent.example.com/authorize?req=eyJ...",
  "expiresAt": "2026-02-01T00:15:00Z"
}

4.4. Token Exchange

After Principal approval, the Authorization Server calls redirectUri?code=AUTH_CODE&state=STATE.

POST /v1/token
Authorization: Bearer <api_key>
Content-Type: application/json

{
  "code": "AUTH_CODE",
  "agentId": "ag_01HXYZ123abc"
}

Response 200 OK:

{
  "grantToken": "eyJhbGciOiJSUzI1NiJ9...",
  "refreshToken": "ref_01HXYZ...",
  "grantId": "grnt_01HXYZ...",
  "scopes": ["calendar:read", "payments:initiate:max_500"],
  "expiresAt": "2026-02-02T00:00:00Z"
}

Refresh tokens are single-use. The Authorization Server MUST rotate the refresh token on every use. Refresh tokens MUST be invalidated when the underlying Grant is revoked.

5. Grant Token Format

5.1. JOSE Header

{
  "alg": "RS256",
  "typ": "JWT",
  "kid": "<key_id>"
}

The alg field MUST be RS256. Authorization Servers MUST NOT issue tokens with any other algorithm. Verifiers MUST explicitly reject tokens with any alg value other than RS256, including none and HS256.

5.2. JWT Claims

{
  "iss": "https://as.example.com",
  "sub": "user_abc123",
  "aud": "https://api.targetservice.com",
  "agt": "did:grantex:ag_01HXYZ123abc",
  "dev": "org_yourcompany",
  "grnt": "grnt_01HXYZ...",
  "scp": ["calendar:read", "payments:initiate:max_500"],
  "iat": 1709000000,
  "exp": 1709086400,
  "jti": "tok_01HXYZ987xyz"
}

The following claims are defined by this specification:

Table 2
Claim Type Required Description
iss string REQUIRED Authorization Server identifier URI
sub string REQUIRED Principal identifier
aud string OPTIONAL Intended audience (target service URI)
agt string REQUIRED Agent DID
dev string REQUIRED Developer organization identifier
grnt string REQUIRED Grant identifier (used for revocation lookup)
scp string[] REQUIRED Array of granted scope strings
iat NumericDate REQUIRED Issued-at time
exp NumericDate REQUIRED Expiration time
jti string REQUIRED Unique token identifier (for replay prevention)

5.3. Token Validation

Services receiving a Grant Token MUST verify all of the following:

  1. The token signature is valid, verified using the JWK Set published at {iss}/.well-known/jwks.json, with the key identified by kid.

  2. The alg header value is RS256. Tokens with any other alg MUST be rejected.

  3. The exp claim has not passed (allowing for a reasonable clock skew of no more than 300 seconds).

  4. If the service has a registered audience identifier, the aud claim matches that identifier.

  5. The scp array contains all scopes required for the requested operation.

  6. For high-stakes operations (see Section 5.4), the token has not been revoked via the online verification endpoint.

5.4. Token Lifetime Guidance

Table 3
Use Case Recommended Maximum TTL
High-stakes actions (payments:initiate, email:send, files:write) 1 hour
Standard agent tasks 8 hours
Long-running background agents 24 hours

Implementations caching revocation state MUST NOT cache for longer than 300 seconds (5 minutes). Services processing high-stakes scopes (payments:initiate, email:send, files:write) SHOULD perform online verification for each token use.

6. Token Revocation

6.1. Revoke a Grant

DELETE /v1/grants/{grantId}
Authorization: Bearer <principal_token>

Effect: all active Grant Tokens issued under this Grant are immediately invalidated. The Grant record is marked revoked with a timestamp.

6.2. Revoke a Specific Token

POST /v1/tokens/revoke
Authorization: Bearer <api_key>
Content-Type: application/json

{
  "jti": "tok_01HXYZ987xyz"
}

Response: 204 No Content.

6.3. Online Verification

POST /v1/tokens/verify
Authorization: Bearer <api_key>
Content-Type: application/json

{
  "token": "eyJhbGciOiJSUzI1NiJ9..."
}

Response 200 OK:

{
  "valid": true,
  "grantId": "grnt_01HXYZ...",
  "scopes": ["calendar:read"],
  "principal": "user_abc123",
  "agent": "did:grantex:ag_01HXYZ123abc",
  "expiresAt": "2026-02-02T00:00:00Z"
}

6.4. JTI Replay Prevention

Authorization Servers MUST track all issued jti values for the lifetime of the corresponding token. If a jti value is presented for verification more than once within its validity window, the Authorization Server MUST return valid: false and SHOULD log an anomaly event.

7. Audit Trail

7.1. Log Entry Schema

{
  "entryId": "alog_01HXYZ...",
  "agentId": "did:grantex:ag_01HXYZ123abc",
  "grantId": "grnt_01HXYZ...",
  "principalId": "user_abc123",
  "developerId": "org_yourcompany",
  "action": "payment.initiated",
  "status": "success",
  "metadata": {
    "amount": 420,
    "currency": "USD",
    "merchant": "Air India"
  },
  "timestamp": "2026-02-01T12:34:56.789Z",
  "hash": "sha256:abc123...",
  "prevHash": "sha256:xyz789..."
}

action values use the format resource.verb (e.g., payment.initiated, email.sent). status MUST be one of success, failure, or blocked.

7.2. Hash Chain

Each entry's hash is computed as a SHA-256 digest [RFC4648] over a canonical representation of the entry:

hash = SHA-256(canonical_json(entry_without_hash) || prevHash)

where canonical_json serializes all fields as a JSON object [RFC8259] with keys sorted alphabetically, and prevHash is the null string for the first entry in a chain. This construction makes any retrospective modification to a historical entry detectable, as it invalidates all subsequent hashes.

7.3. Audit Log Requirements

  • Audit log entries MUST be append-only at the API level. No update or delete endpoints for audit entries are permitted.

  • The Authorization Server MUST reject requests to modify or delete audit entries.

  • The complete audit log for a Grant MUST remain accessible after the Grant is revoked, for a minimum retention period determined by the deployment's compliance requirements.

8. Multi-Agent Delegation

8.1. Delegation Token Claims

When Agent A spawns Agent B, B's Grant Token MUST carry delegation claims linking it to the original Principal's authorization:

{
  "sub": "user_abc123",
  "agt": "did:grantex:ag_B_456",
  "parentAgt": "did:grantex:ag_A_123",
  "parentGrnt": "grnt_parentXYZ",
  "scp": ["email:read"],
  "delegationDepth": 1
}

Additional delegation claims:

Table 4
Claim Type Description
parentAgt string DID of the delegating (parent) Agent
parentGrnt string Grant ID of the parent Grant
delegationDepth integer Number of hops from the root Grant; 0 for root Grants

8.2. Delegation Rules

  • Sub-agent scopes MUST be a strict subset of the parent Grant's scp array. Authorization Servers MUST reject delegation requests whose requested scopes are not fully contained in the parent token's scp claim.

  • delegationDepth MUST be incremented by exactly 1 at each hop.

  • Implementations MUST enforce a developer-configurable delegation depth limit. The RECOMMENDED default limit is 3. Implementations MUST enforce a hard cap of 10 regardless of developer configuration.

  • The expiry of a delegated Grant Token MUST NOT exceed min(parent_token_exp, now + requested_expires_in).

8.3. Delegation Endpoint

POST /v1/grants/delegate
Authorization: Bearer <api_key>
Content-Type: application/json

{
  "parentGrantToken": "eyJhbGciOiJSUzI1NiJ9...",
  "subAgentId": "ag_01HXYZ_sub",
  "scopes": ["email:read"],
  "expiresIn": "1h"
}

The Authorization Server MUST:

  1. Validate that the parent Grant has not been revoked.

  2. Reject with 400 if any requested scope is not present in the parent token's scp claim.

  3. Reject with 400 if the resulting delegationDepth would exceed the configured limit.

  4. Reject with 404 if subAgentId does not belong to the authenticated Developer.

  5. Compute expiry as min(parent token exp, now + expiresIn).

Response 201 Created:

{
  "grantToken": "eyJhbGciOiJSUzI1NiJ9...",
  "grantId": "grnt_01HXYZ_sub",
  "scopes": ["email:read"],
  "expiresAt": "2026-02-01T01:00:00Z"
}

8.4. Cascade Revocation

Revoking a Grant via DELETE /v1/grants/:id MUST atomically revoke all descendant Grants — that is, all Grants whose parent_grant_id traces back to the revoked Grant at any depth. Authorization Servers SHOULD implement this as a single recursive database transaction to eliminate any window during which descendant tokens remain valid.

9. Conformance Requirements

A conformant DAAP Authorization Server MUST expose the following endpoints:

Table 5
Endpoint Description
POST /v1/agents Register an Agent
POST /v1/authorize Initiate the grant authorization flow
POST /v1/token Exchange authorization code for Grant Token
POST /v1/tokens/verify Online token verification
POST /v1/tokens/revoke Revoke a specific token by JTI
GET /v1/grants List a Principal's active Grants
GET /v1/grants/:id Retrieve a single Grant
DELETE /v1/grants/:id Revoke a Grant (cascades to all descendants)
POST /v1/grants/delegate Issue a delegated sub-agent Grant
POST /v1/audit/log Write an audit log entry
GET /v1/audit/entries Query the audit log
GET /v1/audit/:id Retrieve a single audit log entry
GET /.well-known/jwks.json JWK Set for offline token verification
GET /health Health check

The following endpoints are OPTIONAL. Implementations that choose to support an optional extension MUST implement it as specified in this document:

10. Policy Engine

10.1. Purpose

The Policy Engine evaluates developer-defined rules against each authorization request before the consent UI is displayed. Policies enable developers to auto-approve routine low-risk requests and auto-deny requests that violate organizational constraints.

10.2. Effects

Table 6
Effect Description
auto_approve Grant Token issued immediately without showing the consent UI
auto_deny Authorization request rejected immediately with 403 Forbidden

10.3. Condition Fields

Table 7
Field Type Description
scopes string[] Matches when the requested scopes are a subset of this list
principalId string Matches a specific Principal identifier
agentId string Matches a specific Agent identifier
timeWindow object Time constraint: { "startHour": N, "endHour": N, "days": [1..7] } where days are ISO weekday integers (1=Monday, 7=Sunday)

10.4. Evaluation Order

Policy evaluation MUST follow this order:

  1. auto_deny rules are evaluated first. The first matching deny rule wins and the request is rejected immediately.

  2. auto_approve rules are evaluated next. The first matching allow rule causes the Grant Token to be issued.

  3. If no rule matches, the consent UI is displayed to the Principal.

This ordering ensures that restrictive policies cannot be bypassed by a conflicting allow rule.

11. Anomaly Detection

11.1. Purpose

The anomaly detection system monitors Agent behavior at runtime against each Agent's established activity baseline. It identifies behavioral deviations and surfaces them to Developers for review.

11.2. Non-Blocking Requirement

Anomaly detection MUST NOT block token issuance. Detection operates asynchronously as an advisory layer. Authorization Servers MUST NOT delay Grant Token responses pending anomaly analysis.

11.3. Anomaly Types

Table 8
Type Description
unusual_scope_access Agent requested scopes outside its established pattern
high_frequency Token issuance rate significantly exceeds the agent's baseline
off_hours_activity Activity detected outside the Principal's normal active hours
new_principal Agent is requesting access for a previously unserved Principal
cascade_delegation Delegation chain depth approaching or exceeding configured limits

11.4. Severity Levels

Anomaly severity MUST be one of: low, medium, high, critical.

12. Security Considerations

12.1. Algorithm Restrictions

All Grant Tokens MUST be signed with RS256 (RSASSA-PKCS1-v1_5 with SHA-256). Symmetric signing algorithms, including HS256, are NOT PERMITTED. All verifiers MUST explicitly reject tokens presenting alg: none or any symmetric algorithm, regardless of library defaults. This prevents algorithm confusion attacks as described in [RFC8725].

RSA key moduli MUST be at least 2048 bits. Authorization Servers generating or importing signing keys MUST enforce this minimum.

12.2. Token Replay Prevention

Every issued Grant Token carries a unique jti claim. Authorization Servers providing online verification MUST track issued jti values and reject any verification request presenting a jti that has already been used, for the full lifetime of the token.

12.3. CSRF and Redirect URI Security

The state parameter in the authorization request MUST be a cryptographically random, unpredictable value generated per-request. Developer callback handlers MUST validate the returned state value against the value sent in the original request.

Redirect URIs MUST be pre-registered by the Developer for each Agent. Authorization Servers MUST perform exact-match comparison of the redirectUri in each authorization request against the pre-registered set. Prefix matching and wildcard matching are NOT PERMITTED.

12.4. Scope Reduction for Delegation

Delegated Grant Tokens MUST carry a scope set that is a strict subset of the parent Grant's scope set. Authorization Servers MUST enforce this at token issuance time; it MUST NOT be enforced only at verification time.

12.5. Revocation Propagation

Authorization Servers MUST propagate grant revocation to all descendant grants atomically. The maximum allowable latency between a revocation request and the invalidation of all descendant tokens via the online verification endpoint is implementation-defined, but implementations SHOULD target sub-second propagation.

Implementations caching revocation state MUST NOT cache for longer than 300 seconds.

12.7. Audit Log Integrity

The hash-chain construction defined in Section 7 ensures that any modification to a historical audit entry is detectable. Implementations MUST store audit log entries in an append-only manner and MUST expose no API for modification or deletion of audit entries. Audit log export implementations SHOULD verify the hash chain before serving exports.

12.8. Enterprise Identity Security

SSO callback handlers MUST validate both the state (CSRF protection) and nonce (replay protection) parameters before establishing a session. ID tokens received in the SSO callback MUST be cryptographically verified against the identity provider's JWKS endpoint before any claims are trusted.

SCIM provisioning endpoints MUST authenticate via a credential (SCIM Bearer token) that is entirely separate from the Developer API key infrastructure. Compromise of a Developer API key MUST NOT grant access to SCIM provisioning endpoints, and vice versa.

13. IANA Considerations

13.1. JWT Claims Registration

This document requests registration of the following claims in the IANA "JSON Web Token Claims" registry established by [RFC7519]:

agt:

Claim Name: agt

Claim Description: Agent Decentralized Identifier

Change Controller: IETF

Specification Document: This document, Section 5

dev:

Claim Name: dev

Claim Description: Developer organization identifier

Change Controller: IETF

Specification Document: This document, Section 5

grnt:

Claim Name: grnt

Claim Description: Grant identifier for revocation lookup

Change Controller: IETF

Specification Document: This document, Section 5

scp:

Claim Name: scp

Claim Description: Array of granted authorization scope strings

Change Controller: IETF

Specification Document: This document, Section 5

parentAgt:

Claim Name: parentAgt

Claim Description: DID of the delegating parent Agent

Change Controller: IETF

Specification Document: This document, Section 8

parentGrnt:

Claim Name: parentGrnt

Claim Description: Grant identifier of the parent Grant

Change Controller: IETF

Specification Document: This document, Section 8

delegationDepth:

Claim Name: delegationDepth

Claim Description: Number of delegation hops from the root Grant

Change Controller: IETF

Specification Document: This document, Section 8

13.2. Well-Known URI Registration

No new Well-Known URIs are defined by this specification. Implementations use the existing /.well-known/jwks.json path established by [RFC8414].

14. References

14.1. Normative References

[RFC2119]
Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, , <https://www.rfc-editor.org/rfc/rfc2119>.
[RFC3986]
Berners-Lee, T., Fielding, R., and L. Masinter, "Uniform Resource Identifier (URI): Generic Syntax", STD 66, RFC 3986, DOI 10.17487/RFC3986, , <https://www.rfc-editor.org/rfc/rfc3986>.
[RFC4648]
Josefsson, S., "The Base16, Base32, and Base64 Data Encodings", RFC 4648, DOI 10.17487/RFC4648, , <https://www.rfc-editor.org/rfc/rfc4648>.
[RFC6749]
Hardt, D., Ed., "The OAuth 2.0 Authorization Framework", RFC 6749, DOI 10.17487/RFC6749, , <https://www.rfc-editor.org/rfc/rfc6749>.
[RFC6750]
Jones, M. and D. Hardt, "The OAuth 2.0 Authorization Framework: Bearer Token Usage", RFC 6750, DOI 10.17487/RFC6750, , <https://www.rfc-editor.org/rfc/rfc6750>.
[RFC7517]
Jones, M., "JSON Web Key (JWK)", RFC 7517, DOI 10.17487/RFC7517, , <https://www.rfc-editor.org/rfc/rfc7517>.
[RFC7518]
Jones, M., "JSON Web Algorithms (JWA)", RFC 7518, DOI 10.17487/RFC7518, , <https://www.rfc-editor.org/rfc/rfc7518>.
[RFC7519]
Jones, M., Bradley, J., and N. Sakimura, "JSON Web Token (JWT)", RFC 7519, DOI 10.17487/RFC7519, , <https://www.rfc-editor.org/rfc/rfc7519>.
[RFC7636]
Sakimura, N., Ed., Bradley, J., and N. Agarwal, "Proof Key for Code Exchange by OAuth Public Clients", RFC 7636, DOI 10.17487/RFC7636, , <https://www.rfc-editor.org/rfc/rfc7636>.
[RFC8174]
Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, , <https://www.rfc-editor.org/rfc/rfc8174>.
[RFC8259]
Bray, T., Ed., "The JavaScript Object Notation (JSON) Data Interchange Format", STD 90, RFC 8259, DOI 10.17487/RFC8259, , <https://www.rfc-editor.org/rfc/rfc8259>.
[RFC8414]
Jones, M., Sakimura, N., and J. Bradley, "OAuth 2.0 Authorization Server Metadata", RFC 8414, DOI 10.17487/RFC8414, , <https://www.rfc-editor.org/rfc/rfc8414>.

14.2. Informative References

[DID-CORE]
W3C, "Decentralized Identifiers (DIDs) v1.0", , <https://www.w3.org/TR/did-core/>.
[RFC7009]
Lodderstedt, T., Ed., Dronia, S., and M. Scurtescu, "OAuth 2.0 Token Revocation", RFC 7009, DOI 10.17487/RFC7009, , <https://www.rfc-editor.org/rfc/rfc7009>.
[RFC7662]
Richer, J., Ed., "OAuth 2.0 Token Introspection", RFC 7662, DOI 10.17487/RFC7662, , <https://www.rfc-editor.org/rfc/rfc7662>.
[RFC8693]
Jones, M., Nadalin, A., Campbell, B., Ed., Bradley, J., and C. Mortimore, "OAuth 2.0 Token Exchange", RFC 8693, DOI 10.17487/RFC8693, , <https://www.rfc-editor.org/rfc/rfc8693>.
[RFC8725]
Sheffer, Y., Hardt, D., and M. Jones, "JSON Web Token Best Current Practices", BCP 225, RFC 8725, DOI 10.17487/RFC8725, , <https://www.rfc-editor.org/rfc/rfc8725>.
[ULID]
Feerasta, A., "Universally Unique Lexicographically Sortable Identifier", , <https://github.com/ulid/spec>.

Appendix A. Comparison with OAuth 2.0 Extensions

DAAP shares OAuth 2.0's fundamental grant model but differs in the following respects:

versus RFC 6749 (OAuth 2.0): OAuth 2.0 defines a general-purpose delegated authorization framework. DAAP specializes this for AI agents by: adding cryptographic agent identity (DID); defining agent-specific JWT claims (agt, dev, grnt, scp); mandating RS256 exclusively; and adding the delegation, audit, policy, and anomaly detection subsystems.

versus RFC 8693 (Token Exchange): Token Exchange [RFC8693] enables a client to exchange one token for another, including impersonation and delegation use cases. DAAP's delegation model serves a narrower purpose — chaining AI agent sub-authorizations back to a human principal — and adds depth-limiting and cascade revocation semantics not present in RFC 8693.

versus RFC 7662 (Token Introspection): Token Introspection [RFC7662] defines an endpoint for resource servers to query token metadata. DAAP's /v1/tokens/verify endpoint serves a similar purpose but returns DAAP-specific fields (agent, principal, scopes) and is used by agent-side SDKs rather than resource servers.

Acknowledgements

The authors thank the members of the IETF OAuth Working Group for prior art in delegated authorization, and the W3C Decentralized Identifier Working Group for the DID specification that DAAP builds upon for agent identity.

Author's Address

Sanjeev Kumar
Grantex