Agent Identity and Discovery (AID)

Internet-Draft	AID	September 2026
Nemethi	Expires 6 March 2027	[Page]

Abstract

Agent Identity and Discovery (AID) answers one question: given a domain, where is the agent and which protocol should a client speak? An AID client queries a DNS TXT record at the well-known subdomain _agent.<domain> and learns the service endpoint URI, protocol token, authentication hint, and optional metadata for that agent.¶

This document defines the AID v2 (`aid2`) record format, client discovery algorithm, exact-host lookup rules, endpoint-proof (PKA) handshake using Ed25519 HTTP Message Signatures, security requirements, and IANA registrations for the `_agent` DNS node name and the `agent` service name. The legacy `aid1` record format is retained as a compatibility format for clients migrating from earlier deployments. AID is intentionally small; after discovery, protocol-specific mechanisms such as MCP or A2A handle communication and capability negotiation.¶

1. Introduction

Applications that need to locate and connect to an agent often rely on out-of-band configuration, centralized directories, or protocol-specific discovery mechanisms. AID defines a single DNS-based bootstrap point that answers one question: given a domain, where is the agent and which protocol should a client speak?¶

AID uses a TXT record at the well-known DNS name _agent.<domain>. The record is small, versioned, and protocol-agnostic. It tells a client where the agent is located, which protocol token applies, what authentication hint to expect, and whether endpoint proof is available.¶

Zero configuration: a user supplies a domain and the client performs discovery.¶
Decentralized deployment: the protocol uses standard DNS publication.¶
Protocol-agnostic bootstrap: AID identifies the next protocol, rather than replacing it.¶
Stable evolution path: the `_agent` label remains stable across record versions.¶
Endpoint proof: when a public key is published, the endpoint can prove control of the matching private key using RFC 9421 HTTP Message Signatures, optionally bound to the queried domain.¶

This revision updates the specification from the AID v1.2 record format described in [RFC-NEMETHI-AID-00] to the current AID v2 (`aid2`) wire format: the record uses an unpadded base64url Ed25519 JWK `x` value for the endpoint-proof key and an RFC 7638 JWK thumbprint as the HTTP Message Signature `keyid`, removing the `kid`/`i` rotation alias used by `aid1`. It also defines an optional domain-binding extension to the endpoint-proof handshake that lets an endpoint signal consent to serve as the agent for the queried domain. The legacy `aid1` format remains valid during a compatibility window.¶

This document requests two IANA registrations. The RFC 8552 registration covers the deployed TXT-based discovery label `_agent`, while the RFC 6335 service-name-only registration reserves `agent` for possible future SRV-based discovery under the same naming family. Both registrations apply to the `_agent` label across `aid1` and `aid2`; the record-version field, not the DNS node name, distinguishes wire-format revisions.¶

AID has been deployed since mid-2025 with client and server SDKs across six languages. This document is an individual contribution, submitted for consideration alongside other agent-discovery proposals under discussion in the IETF DAWN effort.¶

3. AID Record Format

A provider MUST advertise its agent service by publishing a single TXT record at `_agent.<domain>`. Providers MUST use DNS TXT records for discovery. TXT is the deployed discovery record type because it is widely available across DNS providers and registrars. Future versions of AID may adopt a more structured record type, but the `_agent` label remains stable; see Section 8.¶

3.1. Syntax and Parsing

The record MUST be a single semicolon-delimited string of `key=value` pairs. Clients SHOULD trim leading and trailing whitespace from keys and values. Clients MUST ignore unknown keys unless the key has a known legacy meaning that this specification explicitly rejects for `aid2`.¶

If a DNS server splits the TXT record into multiple 255-octet character strings, the client MUST concatenate them in order before parsing. Providers SHOULD keep total payload size below 255 bytes when possible.¶

Key comparison is case-insensitive. Clients MUST recognize the single-letter lowercase alias for every defined key. A record MUST NOT include both a full key and its alias. Providers SHOULD emit the short-key form for compact DNS deployment.¶

3.2. Defined Keys

version (`v`) Required: Specification version. For AID v2 this MUST be `aid2`. Example: `v=aid2`.¶
uri (`u`) Required: Absolute `https://` URL for a remote agent, `wss://` for WebSocket, or a local locator. Example: `u=https://api.example.com/mcp`.¶
proto (`p`) Required: Protocol token from Section 9.2. Example: `p=mcp`.¶
auth (`a`) Recommended: Authentication hint token from Section 9.1. Example: `a=oauth2_code`.¶
desc (`s`) Optional: Short human-readable description. MUST be no more than 60 UTF-8 bytes. Example: `s=Primary AI Gateway`.¶
docs (`d`) Optional: Absolute `https://` URL for human-readable documentation. Example: `d=https://docs.example.com/agent`.¶
dep (`e`) Optional: ISO 8601 UTC timestamp indicating deprecation. Example: `e=2027-01-01T00:00:00Z`.¶
pka (`k`) Optional: Unpadded base64url Ed25519 public key. The value is exactly the RFC 8037 [RFC8037] JWK `x` member. Example: `k=JrQLj5P_89iXES9-vFgrIy29clF9CC_oPPsw3c5D0bs`.¶

AID v2 records MUST NOT use `kid` or `i` for endpoint proof. A `v=aid2` record containing `kid` or `i` is invalid.¶

3.3. Examples

_agent.example.com. 300 IN TXT (
  "v=aid2;u=https://api.example.com/mcp;"
  "p=mcp;a=pat;s=Example AI Tools"
)

Figure 1: Remote MCP Agent

_agent.example.com. 300 IN TXT (
  "v=aid2;p=mcp;u=https://api.example.com/mcp;"
  "k=JrQLj5P_89iXES9-vFgrIy29clF9CC_oPPsw3c5D0bs;"
  "a=oauth2_code;s=Secure AI Gateway"
)

Figure 2: Remote MCP with PKA

_agent.example.com. 300 IN TXT (
  "v=aid2;p=websocket;u=wss://agent.example.com/session;"
  "a=oauth2_code;s=Streaming Agent"
)

Figure 3: WebSocket Agent

_agent.grafana.com. 300 IN TXT (
  "v=aid2;u=docker:grafana/mcp:latest;"
  "p=local;a=pat;s=Run Grafana agent locally"
)

Figure 4: Local Agent via Docker

_agent.example.com. 300 IN TXT (
  "v=aid1;p=mcp;u=https://api.example.com/mcp;"
  "k=<legacy-v1-multibase-key>;i=g1"
)
_agent.example.com. 300 IN TXT (
  "v=aid2;p=mcp;u=https://api.example.com/mcp;"
  "k=<same-key-as-v2-base64url>"
)

Figure 5: Legacy Compatibility During Migration

Clients that support both versions partition records by version before ambiguity checks. A same-key migration from legacy `aid1` to current `aid2` is not a key rotation when the decoded Ed25519 public key bytes are the same.¶

4. Client Discovery Procedure

4.1. Discovery Algorithm

When an AID client is given a domain, it MUST perform the following steps:¶

Normalize the domain. If the input contains non-ASCII labels, convert them to A-label form as described in [RFC5890].¶
Query the TXT record at `_agent.<exact-host-user-entered>`. The client MUST NOT walk up the DNS hierarchy.¶
Parse the TXT answer set as semicolon-delimited `key=value` records. Key comparison MUST be case-insensitive, and clients MUST recognize single-letter aliases for all defined keys.¶
Validate records using versioned rules. For `aid2`, require `v`, `u`, and `p`; reject a record containing both a full key and its alias; reject `kid` or `i`; and reject malformed `k` values.¶
Partition valid records by AID major version.¶
Select the highest supported valid version allowed by local policy. Clients that support `aid2` SHOULD prefer `aid2` over `aid1`. Within the selected version, if exactly one valid record exists, use it. If more than one valid record exists, the client MUST fail due to ambiguity; clients MUST NOT choose among multiple valid same-version records by DNS answer order.¶
If `docs` is present, the client MAY display it. If `dep` is in the future, the client SHOULD warn. If `dep` is in the past, the client SHOULD fail gracefully.¶
If `k` is present, the client MUST perform the endpoint-proof procedure in Appendix A.¶
For `aid2` records, clients SHOULD request domain binding by sending the `AID-Domain` header (the A-label, lowercased, portless queried host from step 1) as described in Appendix A.7, unless local policy disables it (Section 5.3, `domain-binding=off`).¶
Return the discovered endpoint, protocol, metadata, PKA state (including the domain-binding indicator when domain binding was requested), and trust source.¶

Malformed answers do not matter when there is exactly one valid record in the selected version. Returning clients that previously selected `aid2` SHOULD treat an `aid1`-only result as a version downgrade.¶

If no DNS record is found or the DNS lookup fails, the client MAY attempt the `.well-known` fallback on the same exact host; see Appendix B. If both DNS and fallback fail, discovery fails.¶

4.2. Standard Client Error Codes

Client implementations SHOULD use these codes to report specific failure modes.¶

Table 1: Client Error Codes
Code	Name	Meaning
1000	ERR_NO_RECORD	No AID DNS record was found for the domain.
1001	ERR_INVALID_TXT	A record was found but is malformed, invalid, or ambiguous.
1002	ERR_UNSUPPORTED_PROTO	The record is valid, but the client does not support the selected protocol.
1003	ERR_SECURITY	Discovery failed due to a security policy or failed endpoint proof.
1004	ERR_DNS_LOOKUP_FAILED	The DNS query failed for a network-related reason.
1005	ERR_FALLBACK_FAILED	The `.well-known` fallback failed or returned invalid data.

4.3. Exact-Host Semantics and Explicit Delegation

Discovery is exact-host by default. If the application asks for `app.team.example.com`, the canonical AID query name is `_agent.app.team.example.com`.¶

Clients MUST NOT walk up the DNS hierarchy looking for `_agent.team.example.com` or `_agent.example.com`. If an operator wants a child host to inherit a shared record, that inheritance MUST be expressed in DNS for the exact queried name, for example by using a `CNAME` at the child host's `_agent` label.¶

_agent.app.team.example.com. 300 IN CNAME (
  _agent.shared.team.example.com.
)
_agent.shared.team.example.com. 300 IN TXT (
  "v=aid2;p=mcp;u=https://gateway.team.example.com/mcp;"
  "k=<current-key>"
)

Figure 6: Exact-Host and Delegated Examples

The canonical v2 location is the base record `_agent.<domain>`. Providers MAY additionally publish protocol-specific names such as `_agent._mcp.<domain>` or `_agent._a2a.<domain>` for legacy clients, diagnostics, or explicitly configured base-failure probing. Protocol-specific names, when used, always use the underscore form `_agent._<proto>.<domain>`; clients MUST NOT perform a compatibility lookup at `_agent.<proto>.<domain>`.¶

When an application explicitly requests a protocol, v2 clients still query the canonical base name `_agent.<exact-host-user-entered>` first and filter that record for the requested protocol. Protocol-prefixed probing is legacy, diagnostic, or base-failure-only behavior where supported and explicitly configured. Clients MUST NOT query parent hosts implicitly in either lookup form.¶

5. Security Considerations

AID publishes public discovery metadata. The TXT record MUST NOT contain secrets. The protocol depends on DNS integrity, TLS for remote endpoints, and explicit client safeguards for local execution.¶

Providers SHOULD sign AID records with DNSSEC when available. Clients SHOULD perform DNSSEC validation when DNSSEC-signed answers are available.¶
A remote agent's `uri` MUST use `https://`, except `proto=websocket` which MUST use `wss://`. Clients MUST perform standard TLS certificate and hostname validation as described in [RFC9525].¶
When `k` is present, clients MUST verify endpoint proof as described in Appendix A.¶
Clients that support `proto=local` MUST implement the safeguards described below.¶
Clients MUST NOT automatically follow cross-origin redirects from a discovered endpoint without policy approval or user confirmation.¶

Clients that support `proto=local` MUST implement the following safeguards:¶

Explicit consent: before first execution, the client MUST display the full resolved command and require explicit confirmation.¶
Integrity check: the client MUST compute and cache a cryptographic fingerprint of the `uri` and `proto` values. If these values change on a later lookup, the client MUST re-trigger the full consent process.¶
No shell interpretation: arguments derived from the `uri` MUST be passed atomically to the underlying OS execution call to prevent command injection.¶
No nested discovery: the client MUST reject a `local` execution `uri` that could be interpreted as a command that initiates another AID discovery request.¶
Sandboxing: clients SHOULD run local agents in a sandboxed environment with minimum necessary permissions.¶

If an initial request to the discovered `uri` returns an HTTP redirect (`301`, `302`, `307`, or `308`) to a different origin, clients SHOULD treat this as a potential security risk. Clients MUST NOT follow such cross-origin redirects automatically. Implementations MAY either terminate with `ERR_SECURITY` or require explicit user confirmation before proceeding. Clients MUST NOT follow redirects during PKA verification or `.well-known` fallback.¶

5.1. What PKA Proves

PKA proves exactly this: the endpoint reached at the discovered URI controls the Ed25519 private key corresponding to the public key currently published in the domain's selected AID record.¶

PKA does not prove: that a user authorized a specific action; that an OAuth token is valid; that a SPIFFE SVID belongs to a trust domain; that an internal policy engine approved a request; that a key change is cryptographically continuous with a previous key; or that the endpoint consents to serve as the agent for the queried domain.¶

Because the response signature binds only the endpoint's own request context, any domain can publish a record containing another operator's endpoint URI and public key, and the endpoint proof still verifies. This unauthorized-association gap does not let the publishing domain impersonate the endpoint, but it falsely implies a relationship between the domain and the endpoint. Clients that need the endpoint's consent to the association use the domain-binding profile in Appendix A.7; v2 clients SHOULD request this by default.¶

5.2. Threat Model

AID's security model addresses the following threat landscape.¶

Assumptions:¶

DNS resolvers are trusted for transport unless a client requires DNSSEC.¶
HTTPS endpoints are verified through standard TLS certificate validation.¶
The TXT record is public data; no secrets are transmitted through DNS.¶

Mitigations provided:¶

DNS spoofing or cache poisoning: DNSSEC validation, when available.¶
Endpoint impersonation: PKA endpoint proof with Ed25519 HTTP Message Signatures.¶
Unauthorized association: the domain-binding profile (Appendix A.7), which v2 clients SHOULD request by default. This mitigation applies only when a client requires domain binding; merely requesting it does not stop an attacker-controlled endpoint from returning a valid unbound proof.¶
PKA removal or key replacement: returning clients can detect changes when they retain previous security state.¶
Version downgrade: returning clients can detect `aid2` to `aid1` downgrade when they retain previous version state.¶
Command injection in local agents: local execution safeguards.¶
Cross-origin redirects: PKA redirects are rejected, and cross-origin protocol redirects are not automatic.¶

Explicitly out of scope:¶

Compromised authoritative DNS servers beyond DNSSEC.¶
Active attackers after TLS validation fails.¶
Authorization, delegation, user consent, reputation, or workload federation.¶
Managed cryptographic rotation in the core DNS record.¶

5.3. Enterprise Policy Modes

Clients that expose enterprise controls SHOULD provide policy presets and MAY expose the underlying policy knobs directly.¶

Table 2: Normative Policy Presets
Preset	PKA	DNSSEC	Well-known	Downgrade	Domain binding
balanced	if-present	prefer	auto	warn	prefer
strict	require	require	disable	fail	require

The underlying policy knobs are:¶

PKA policy: `if-present | require`¶
DNSSEC policy: `off | prefer | require`¶
Well-known policy: `auto | disable`¶
Downgrade policy: `off | warn | fail`¶
Domain-binding policy: `off | prefer | require`¶

Policy semantics are as follows:¶

PKA `require`: discovery MUST fail with `ERR_SECURITY` if the selected record does not publish `k`.¶
DNSSEC `prefer`: clients SHOULD continue when DNSSEC cannot be validated, but SHOULD surface a warning.¶
DNSSEC `require`: clients MUST fail with `ERR_SECURITY` when DNSSEC validation is unavailable or unsuccessful for the selected DNS answer.¶
Well-known `disable`: clients MUST NOT use `/.well-known/agent` fallback.¶
Downgrade `warn`: if a previously seen `k` disappears or changes, clients SHOULD surface a warning. This also applies to an `aid2` to `aid1` downgrade.¶
Downgrade `fail`: if a previously seen `k` disappears or changes, clients MUST fail with `ERR_SECURITY`. This also applies to an `aid2` to `aid1` downgrade.¶
Domain-binding `off`: the client does not send `AID-Domain` on PKA requests.¶
Domain-binding `prefer` (default): the client sends `AID-Domain`. A domain-bound proof (one whose covered set includes `"aid-domain";req`) is recorded as such; an unbound proof is still accepted. `prefer` records the outcome but does not enforce it.¶
Domain-binding `require`: when an endpoint proof is performed for a record containing `k`, discovery fails unless the proof is domain-bound. This is the only mode that mitigates unauthorized association (Section 5.1, Appendix A.7); merely sending `AID-Domain` does not. Has no effect when no `k` is present or PKA yields no proof. `pka=require` and `domain-binding=require` compose: `pka=require` fails first when `k` is absent, then `domain-binding=require` enforces binding on the resulting proof.¶

When a record is discovered through the `.well-known` fallback (`trustSource=well-known-tls`; see Appendix B), the queried host and the TLS-validated host are the same origin, so an `AID-Domain` binding there is largely redundant with TLS host validation. Clients still send `AID-Domain` and `domain-binding=require` still enforces, but the binding adds little beyond TLS in that path.¶

If discovery succeeds only through `.well-known`, the result cannot satisfy `dnssec=require`.¶

7. Rotation Stance

AID v2 core does not define DNS-level cryptographic key rotation. The core record says, for example:¶

_agent.acme.com TXT (
  "v=aid2;p=mcp;u=https://agent.acme.com/mcp;"
  "k=<current-key>"
)

That means:¶

DNS currently publishes this endpoint and this key.¶
The endpoint can prove possession of the corresponding private key.¶
Verifiers can derive stable key identity from `k`.¶

It does not mean:¶

The new key is authorized by the old key.¶
Multiple keys are active through DNS.¶
DNS provides validity windows.¶
AID core provides request provenance or delegated signing infrastructure.¶

If the key changes, the key changed. Clients with previous state decide whether to warn, fail, or accept according to local policy (Section 5.3).¶

7.1. Why No Multi-Key RRset in Core

A shape such as the following is not valid in v2 core:¶

_agent.acme.com. 300 IN TXT (
  "v=aid2;p=mcp;u=https://agent.acme.com/mcp;"
  "k=<old-key>"
)
_agent.acme.com. 300 IN TXT (
  "v=aid2;p=mcp;u=https://agent.acme.com/mcp;"
  "k=<new-key>"
)

It changes the model from a domain resolving to one selected endpoint record with an optional proof key, to a domain resolving to one selected endpoint descriptor with an active key set from which a response `keyid` selects a key. That requires rules for normalized non-key field equivalence, duplicate handling, partial DNS propagation, key-set pinning, downgrade policy, SDK return types, and conformance fixtures. It also does not solve lost-key or compromised-key recovery.¶

7.2. Future Rotation Profile

If AID later needs managed rotation for pinned clients or provenance profiles, that should be a separate HTTP key-directory profile, likely using JWKS or Web Bot Auth style overlap:¶

{
  "keys": [
    {
      "kty": "OKP",
      "crv": "Ed25519",
      "kid": "<jwk-thumbprint>",
      "x": "<public-key>",
      "use": "sig",
      "nbf": 1712793600,
      "exp": 1715385600
    }
  ]
}

If such a directory chains to DNS `k`, that chaining is the defining property of an AID-anchored key directory. If it does not chain to DNS `k`, it is a normal external key directory outside AID core.¶

Table 3: Protocol Tokens
Token	Meaning	Allowed `uri` scheme(s)
mcp	Model Context Protocol	https://
a2a	Agent-to-Agent Protocol	https://
openapi	OpenAPI specification document	https://
grpc	gRPC over HTTP/2 or HTTP/3	https://
graphql	GraphQL over HTTP	https://
websocket	WebSocket transport	wss://
local	The agent runs locally on the client machine	docker:, npx:, pip:
zeroconf	mDNS or DNS-SD service discovery	zeroconf:<service_type>
ucp	Universal Commerce Protocol	https://

10. IANA Considerations

This document requests registration in two IANA registries. The registrations serve different purposes. The RFC 8552 request covers AID TXT discovery at `_agent.<domain>`, applicable across the `aid1` and `aid2` record versions defined in this document. The RFC 6335 request reserves the service name `agent` for potential future DNS service discovery usage under `_agent._tcp.<domain>`.¶

10.1. Registration in the Underscored and Globally Scoped DNS Node Names Registry

IANA is requested to register the following entry in the "Underscored and Globally Scoped DNS Node Names" registry established by [RFC8552].¶

Table 4: Requested RFC 8552 Registration
RR Type	TXT
Node Name	_agent
Reference	This document

The `_agent` node name is used exclusively for Agent Identity and Discovery. A single TXT record published at `_agent.<domain>` contains semicolon-delimited key/value pairs that identify an agent endpoint, protocol token, and optional metadata as defined in Section 3 and Section 4.¶

The requested node name is specific to AID and does not reserve the broader concept of agents or agent-related discovery generally. Protocol-specific labels of the form `_agent._<proto>.<domain>` are subordinate names beneath the registered `_agent` node and do not require separate global registration.¶

10.2. Registration in the Service Name and Transport Protocol Port Number Registry

IANA is requested to register the following service-name-only entry in the "Service Name and Transport Protocol Port Number" registry defined by [RFC6335].¶

Table 5: Requested RFC 6335 Registration
Service Name	agent
Transport Protocol(s)	tcp
Description	Agent Identity and Discovery (AID): DNS-based discovery of agent service endpoints
Port Number	N/A
Assignment Notes	No port number is requested. This is a service-name-only registration intended to establish `agent` for possible future SRV-based discovery under `_agent._tcp.<domain>` while preserving the stable `_agent` label.
Reference	This document

Appendix A. PKA Handshake

When `k` is present, clients MUST verify endpoint proof using HTTP Message Signatures with Ed25519.¶

A.1. Key Decoding

For v2, `k` MUST be unpadded base64url. Decoding MUST produce exactly 32 octets. Legacy `z...` multibase keys MUST NOT be accepted in `v=aid2`. The corresponding JWK is:¶

{ "kty": "OKP", "crv": "Ed25519", "x": "<k>" }

A.2. Derived `keyid`

The expected HTTP Message Signature `keyid` is the RFC 7638 [RFC7638] JWK thumbprint using SHA-256 over this exact UTF-8 JSON serialization, with no extra spaces:¶

{"crv":"Ed25519","kty":"OKP","x":"<k>"}

The SHA-256 digest is encoded as unpadded base64url. Implementations MUST NOT hash the raw public key bytes directly for `keyid`.¶

A.3. Request and Response Shape

Validated RFC 9421 [RFC9421] Structured Fields shape:¶

Accept-Signature: aid-pka=("@method";req "@target-uri";req
  "@authority";req "@status");created;expires;
  keyid="<jwk-thumbprint>";alg="ed25519";
  nonce="<client-challenge>";tag="aid-pka-v2"
Signature-Input: aid-pka=("@method";req "@target-uri";req
  "@authority";req "@status");created=<unix>;expires=<unix>;
  keyid="<jwk-thumbprint>";alg="ed25519";
  nonce="<client-challenge>";tag="aid-pka-v2"
Signature: aid-pka=:<base64-signature>:
Cache-Control: no-store

The signature MUST NOT cover HTTP `Date`.¶

The client challenge MUST contain at least 32 bytes of entropy and SHOULD be transported as unpadded base64url in the RFC 9421 `nonce` signature parameter. The verifier MUST compare the received `nonce` exactly to the challenge it sent. Servers are not required to store nonce state in v2 core because the verifier supplies the one-shot nonce and the signed response is not cacheable.¶

`created` and `expires` are mandatory. `expires` MUST be greater than `created`. `expires` minus `created` MUST NOT exceed 300 seconds and SHOULD be 60 seconds or less. Verifiers MAY allow a small clock-skew tolerance when evaluating `created` and `expires`.¶

Signers MUST emit `alg="ed25519"` lowercase. Verifiers MUST compare the semantic algorithm value case-insensitively and MUST reconstruct `@signature-params` from the received Structured Field value.¶

A.4. Covered Components

The base v2 PKA response signature covers `"@method";req`, `"@target-uri";req`, `"@authority";req`, and `"@status"`. `@method`, `@target-uri`, and `@authority` are request-derived components and therefore use `;req`. `@status` is response-derived and does not use `;req`. The domain-binding profile in Appendix A.7 extends this base covered set with `"aid-domain";req`; the tag remains `aid-pka-v2`.¶

`@status` signs the status actually returned. PKA does not require status `200`. A signed `401` can still prove endpoint authenticity before an OAuth or auth.md handoff continues.¶

A.5. URI, Authority, and Redirects

Clients MUST NOT follow redirects during PKA verification. The request context is the discovered endpoint URI after fragment removal. Query strings are preserved.¶

`@authority` uses the externally visible request authority: lowercase hostname, omit the default port, retain a non-default port. Servers behind reverse proxies must sign the externally visible request context, not internal hop-local scheme, host, or port values.¶

A.6. Verifier Summary

A verifier accepts a v2 PKA response only when:¶

The selected AID record contains a valid v2 `k`.¶
The response contains a valid `Signature-Input` and `Signature`.¶
The tag is `tag="aid-pka-v2"` and the covered components are either exactly the four components in Appendix A.4, or those four plus the optional `"aid-domain";req` component (between `"@authority";req` and `"@status"`) per the domain-binding profile in Appendix A.7.¶
`keyid` equals the RFC 7638 thumbprint derived from DNS `k`.¶
`alg` has semantic value `ed25519`.¶
`nonce` exactly equals the verifier-generated challenge.¶
`created` and `expires` pass freshness checks.¶
The response includes `Cache-Control: no-store`.¶
Ed25519 verification succeeds over the reconstructed RFC 9421 signature base.¶

A.7. Domain Binding

This profile lets an endpoint prove that it consents to serve as the agent for the queried domain, addressing the unauthorized-association gap described in Section 5.1. When `k` is present in an `aid2` record, clients SHOULD request domain binding by default (see Section 4.1). The domain-binding indicator in the discovery result is the expected outcome for well-configured v2 deployments; clients that need a hard enforcement boundary use `domain-binding=require` in enterprise policy (Section 5.3).¶

There is a single RFC 9421 tag for all v2 PKA proofs, `aid-pka-v2`. A proof is domain-bound if and only if its signed covered set includes the `"aid-domain";req` component, positioned strictly between `"@authority";req` and `"@status"`. The covered set is part of `@signature-params`, which is itself signed, so coverage is authenticated and cannot be altered without invalidating the signature. The `tag` identifies the application profile per [RFC9421] Section 2.3; it is the covered set, not a separate tag, that distinguishes an unbound proof from a domain-bound one.¶

A client requesting domain binding sends the queried domain in the `AID-Domain` request header and requests an extended response signature whose covered set adds `"aid-domain";req`:¶

AID-Domain: example.com
Accept-Signature: aid-pka=("@method";req "@target-uri";req
  "@authority";req "aid-domain";req "@status");created;expires;
  keyid="<jwk-thumbprint>";alg="ed25519";
  nonce="<client-challenge>";tag="aid-pka-v2"

The `AID-Domain` value is the exact host the client queried in Section 4.1 (step 2), normalized to its A-label form per Section 4.1 (step 1), lowercased, and without a trailing dot or port. Signer and verifier MUST use this identical byte value.¶

A server that supports this profile and serves the named domain responds with the Appendix A.3 shape, except that the covered set includes `"aid-domain";req` after `"@authority";req`. The `aid-domain` component is the request header field; its RFC 9421 component identifier is the lowercased field name `aid-domain`, so the signature binds the exact value the client sent.¶

A server that supports this profile but does not serve the named domain MUST NOT produce a signature covering that `AID-Domain` value. It SHOULD respond with status `403` and no `Signature-Input` header. A server that does not support this profile ignores the header and responds with the base Appendix A.3 shape, whose covered set omits `aid-domain`, which remains a valid endpoint proof without domain binding.¶

A response that carries no valid `aid-pka-v2` signature, including a `403` refusal, is a failed endpoint proof under Section 5, so discovery fails whenever the selected record contains `k`. A profile-aware endpoint that refuses an unserved domain therefore causes discovery to fail for that domain, which is the intended outcome.¶

Verifier rules, in addition to Appendix A.6:¶

The covered set of an `aid-pka-v2` response MUST be exactly the four base components of Appendix A.3, or those four base components plus `"aid-domain";req` (between `"@authority";req` and `"@status"`) and nothing else. A response whose covered set includes `aid-domain` is domain-bound; one whose covered set omits it is unbound.¶
A response is verified as domain-bound only when its covered set includes `"aid-domain";req`. A client that sent an `AID-Domain` request header rebuilds the signature base using its own canonical queried domain as the `aid-domain` value; a domain-bound proof verifies only when the signed domain matches that value.¶
A client that did not send `AID-Domain` MUST reject a response whose covered set includes `aid-domain` (fail-closed).¶
Clients that request domain binding MUST expose a boolean domain-binding indicator as part of PKA state, set `true` only when `aid-domain` was covered and the proof verified for the queried domain, and `false` for a verified unbound proof.¶
Requesting domain binding does not by itself mitigate unauthorized association, because an attacker-controlled endpoint can ignore `AID-Domain` and return a valid unbound proof. A client gains the mitigation only when it requires domain binding and rejects unbound proofs by local policy.¶

Domain binding is a statement by the endpoint that it serves the named domain. It does not prove authorization, delegation, or organizational identity.¶

In AID v2, domain binding is optional-but-default: clients SHOULD send `AID-Domain`, and an unbound proof (one whose covered set omits `aid-domain`) remains a valid outcome unless local policy requires binding (`domain-binding=require`; see Section 5.3). A future major version (`aid3`) is expected to make sending `AID-Domain` REQUIRED for clients performing PKA and to make rejecting unbound proofs the baseline. Establishing high adoption while the installed base is small is intended to minimize switching cost at that transition; implementations that already send `AID-Domain` by default require no change at `aid3`.¶

Composition Notes

This appendix is non-normative. It describes how AID composes with adjacent mechanisms without incorporating them into AID core.¶

Web Bot Auth

AID v2 follows Web Bot Auth where the layers match: Ed25519, RFC 9421 HTTP Message Signatures, RFC 7638 JWK thumbprints, `created`, `expires`, `nonce`, and `tag`.¶

AID does not become Web Bot Auth. Web Bot Auth signs automated client requests to origins. AID PKA proves endpoint control for a DNS-discovered agent endpoint.¶

Operators may reuse key material across AID and Web Bot Auth if their threat model allows it, but this document does not recommend reuse. Reuse shares blast radius between endpoint proof and request-signing.¶

Authorization and OAuth

AID can support an authorization handoff without adding authorization-specific fields to core:¶

AID resolves a domain to `u` and `p`.¶
If `k` is present, PKA verifies the endpoint.¶
The client follows the endpoint's protocol and authorization layer, including RFC 9728 [RFC9728] OAuth 2.0 Protected Resource Metadata and RFC 8414 [RFC8414] OAuth 2.0 Authorization Server Metadata where present.¶

AID does not specify audience restriction, provider trust lists, registration payloads, scopes, credential types, revocation, or an authorization metadata registry.¶

SPIFFE and WIMSE

AID may be used by future SPIFFE or WIMSE profiles as a public first-contact anchor.¶

AID v2 core does not define:¶

SPIFFE trust-domain mapping;¶
SPIFFE bundle federation;¶
WIMSE hop re-binding;¶
OAuth client registration;¶
workload authorization.¶

Relationship to Key-Addressed Identity Models

AID v2 remains compatible with the general idea of compact Ed25519 key material as used by some key-addressed identity designs, but it does not adopt a key-addressed identity model. In such designs, the public key itself is the address and DNS-shaped records are signed directly under that key. AID remains DNS-authority-rooted: the DNS owner publishes the current endpoint and current endpoint-proof key inside a TXT payload at a name derived from the domain, not from the key.¶

Specification and reviewer-facing language should say "DNS-current endpoint/key" and should not imply self-certifying names, key-addressed identity, or cryptographic continuity across key changes.¶