| Internet-Draft | AID | June 2026 |
| Nemethi | Expires 3 December 2026 | [Page] |
Agent Identity and Discovery (AID) is a minimal, DNS-first discovery protocol for locating agent and tool endpoints. Given a domain name, an AID client queries a DNS TXT record at the well-known base name _agent.<domain> and learns the endpoint URI, protocol token, authentication hint, and optional endpoint-proof key for that domain.¶
AID v2 is the current wire format defined by this document. It uses `v=aid2`, keeps _agent.<domain> as the canonical discovery name, and updates endpoint proof so that the DNS key is an Ed25519 JSON Web Key (JWK) `x` value and the HTTP signature `keyid` is derived as an RFC 7638 JWK thumbprint. Legacy `aid1` records remain a compatibility format for clients that need to read older deployments.¶
AID is intentionally a first-contact discovery layer. After discovery, protocol-specific mechanisms such as MCP, A2A, OpenAPI, OAuth, auth.md, or other application layers handle communication, capability negotiation, authentication, authorization, and policy.¶
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 3 December 2026.¶
Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
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 DNS-based bootstrap point that answers one question: given a domain, where is the agent and which protocol should a client speak?¶
AID resolves a domain to the current agent endpoint, protocol hint, and optional endpoint-proof key. After AID discovery, the selected endpoint's protocol, authorization, registration, identity, and policy layers take over.¶
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.¶
This document specifies the wire format and client behavior for AID v2. AID v2 keeps the stable `_agent` label and TXT payload model while changing the PKA key encoding and endpoint proof profile. The label governance question is independent of the v2 key-format cleanup; see Section 9.¶
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.¶
A provider MUST advertise its agent service by publishing a DNS TXT record at the canonical base name `_agent.<domain>`.¶
For AID v2, providers MUST use DNS TXT records for discovery. TXT records remain the baseline because they are widely deployable across DNS providers and registrar control panels. Future AID versions may adopt other record types, but v2 uses the stable `_agent` label and TXT payload defined here.¶
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 the total payload below 255 octets when practical.¶
Clients MUST recognize single-letter lowercase aliases. A record MUST NOT include both a full key and its alias. Key comparisons are case-insensitive. Providers SHOULD emit the short-key form for compact DNS deployment.¶
| Key | Alias | Requirement | Description | Example |
|---|---|---|---|---|
| version | v | Required | Specification version. For v2 this MUST be `aid2`. | `v=aid2` |
| uri | u | Required | Absolute `https://` URL for a remote agent, `wss://` for WebSocket, or a local locator. | `u=https://api.example.com/mcp` |
| proto | p | Required | Protocol token from Appendix B. | `p=mcp` |
| auth | a | Recommended | Authentication hint token from Appendix A. | `a=oauth2_code` |
| desc | s | Optional | Short human-readable display text. | `s=Primary AI Gateway` |
| docs | d | Optional | Absolute `https://` URL to human-readable documentation. | `d=https://docs.example.com/agent` |
| dep | e | Optional | ISO 8601 UTC deprecation timestamp. | `e=2027-01-01T00:00:00Z` |
| pka | k | Optional | Unpadded base64url Ed25519 public key. The value is exactly the RFC 8037 JWK `x` member. | `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. The expected key identifier is derived from `k` as described in Appendix D.¶
_agent.example.com. 300 IN TXT "v=aid2;u=https://api.example.com/mcp;p=mcp;a=pat;s=Example AI Tools"
_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"
_agent.example.com. 300 IN TXT "v=aid2;p=websocket;u=wss://agent.example.com/session;a=oauth2_code;s=Streaming Agent"
_agent.grafana.com. 300 IN TXT "v=aid2;u=docker:grafana/mcp:latest;p=local;a=pat;s=Run Grafana agent locally"
_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>"
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.¶
When an AID client is given a domain, it MUST perform the following steps:¶
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 DNS lookup fails, the client MAY attempt the `.well-known` fallback on the same exact host as described in Appendix E. If both DNS and fallback fail, discovery fails.¶
| Code | Name | Meaning |
|---|---|---|
| 1000 | ERR_NO_RECORD | No AID DNS record was found. |
| 1001 | ERR_INVALID_TXT | A record was malformed, invalid, or ambiguous. |
| 1002 | ERR_UNSUPPORTED_PROTO | The protocol token is unsupported. |
| 1003 | ERR_SECURITY | Discovery failed due to a security policy or failed endpoint proof. |
| 1004 | ERR_DNS_LOOKUP_FAILED | The DNS lookup failed for a network-related reason. |
| 1005 | ERR_FALLBACK_FAILED | The `.well-known` fallback failed or returned invalid data. |
Discovery is exact-host by default. If the application asks for `app.team.example.com`, the base query is `_agent.app.team.example.com`.¶
Clients MUST NOT implicitly retry parent hosts such as `_agent.team.example.com` or `_agent.example.com`.¶
If an operator wants child hosts to inherit a shared record, that inheritance MUST be expressed in DNS for the exact queried name. A common pattern is a CNAME from the child host's `_agent` label to a shared `_agent` record.¶
_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>"
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.¶
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 `_agent._<proto>.<exact-host-user-entered>` probing is legacy, diagnostic, or base-failure-only behavior where supported and explicitly configured.¶
Clients MUST NOT perform a compatibility lookup at `_agent.<proto>.<domain>`. Protocol-specific names, when used, always use the underscore form `_agent._<proto>.<domain>`.¶
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.¶
Clients that support `proto=local` MUST implement the following safeguards:¶
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 workload identity belongs to a trust domain, that an internal policy engine approved a request, or that a key change is cryptographically continuous with a previous key.¶
AID addresses DNS spoofing and cache poisoning with optional DNSSEC validation, endpoint impersonation with PKA endpoint proof using Ed25519 HTTP Message Signatures, downgrade detection with remembered state, command injection risks for local execution, and cross-origin redirect abuse for remote endpoints.¶
Returning clients can detect PKA removal, key replacement, and `aid2` to `aid1` version downgrade when they retain previous security state. If the key changes, the key changed; AID v2 core does not define DNS-level cryptographic continuity across key changes.¶
Compromised authoritative DNS servers, active attackers after TLS validation fails, authorization, delegation, user consent, reputation, workload federation, and managed cryptographic rotation in the core DNS record are outside the direct scope of AID v2.¶
Clients that expose enterprise controls SHOULD provide simple policy knobs.¶
Policy semantics are as follows:¶
If discovery succeeds only through `.well-known`, that result cannot satisfy a policy that requires DNSSEC validation.¶
Providers SHOULD publish `_agent` TXT records with a TTL between 300 and 900 seconds. Clients MUST respect the received TTL and MUST NOT cache DNS records longer than that.¶
PKA responses are separate from DNS records. A nonce-bound PKA response MUST include `Cache-Control: no-store`. Clients SHOULD also send `Cache-Control: no-store` on PKA requests.¶
AID v2 core does not define DNS-level cryptographic key rotation. The core record says that DNS currently publishes this endpoint and this key, the endpoint can prove possession of the corresponding private key, and verifiers can derive stable key identity from `k`.¶
_agent.acme.com. 300 IN TXT "v=aid2;p=mcp;u=https://agent.acme.com/mcp;k=<current-key>"
This does not mean that the new key is authorized by the old key, that multiple keys are active through DNS, that DNS provides validity windows, or that 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.¶
AID v2 core rejects the following shape as a core discovery result:¶
_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>"
This changes the model from one selected endpoint record with an optional proof key into one endpoint descriptor with an active key set where the HTTP response `keyid` selects a key. That larger model 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.¶
If AID later needs managed rotation for pinned clients or provenance profiles, it 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": "<publicKey>",
"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.¶
The auth and protocol registries remain compatible with AID v1.2 unless changed through the normal extension process.¶
This revision does not request new IANA actions.¶
AID v2 keeps `_agent.<domain>` as the discovery label. The earlier RFC 8552 `_agent` registration request was rejected on procedural grounds in April 2026. That label-governance question is decoupled from the v2 PKA cleanup. A later working group or BoF process may revisit the label.¶
All authentication tokens are case-sensitive and use lowercase ASCII.¶
The `auth` value is an advisory hint. It does not define registration payloads, credential formats, authorization-server semantics, or trust policy.¶
All protocol tokens are case-sensitive and use lowercase ASCII.¶
| 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:// |
For cross-language SDK consistency, clients SHOULD use the following numeric constants.¶
| Constant | Value |
|---|---|
| ERR_NO_RECORD | 1000 |
| ERR_INVALID_TXT | 1001 |
| ERR_UNSUPPORTED_PROTO | 1002 |
| ERR_SECURITY | 1003 |
| ERR_DNS_LOOKUP_FAILED | 1004 |
| ERR_FALLBACK_FAILED | 1005 |
When `k` is present, clients MUST verify endpoint proof using HTTP Message Signatures with Ed25519 as specified in [RFC9421].¶
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>" }
The expected HTTP Message Signature `keyid` is the RFC 7638 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`.¶
The validated RFC 9421 Structured Fields shape is:¶
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 - 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.¶
The v2 PKA response signature covers:¶
`@method`, `@target-uri`, and `@authority` are request-derived components and therefore use `;req`. `@status` is response-derived and does not use `;req`.¶
`@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 verifier accepts a v2 PKA response only when:¶
AID remains DNS-first. The `.well-known` fallback is a convenience for environments where DNS TXT record creation is restricted.¶
DNS-discovered records have `trustSource=dns`. Fallback-discovered records have `trustSource=well-known-tls`.¶
If a policy requires DNSSEC-backed trust, `.well-known` cannot satisfy it. The `.well-known` URI convention itself is described in [RFC8615].¶
This section is non-normative.¶
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 AID does not recommend reuse. Reuse shares blast radius between endpoint proof and request-signing.¶
AID can support auth.md without adding auth.md-specific fields to core:¶
AID does not specify ID-JAG `aud`, provider trust lists, registration payloads, scopes, credential types, revocation, or auth.md metadata registries.¶
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, or workload authorization.¶
AID v2 remains compatible with the Pkarr-inspired idea of compact Ed25519 key material, but it does not adopt Pkarr's identity model.¶
Pkarr makes the public key the address and signs DNS packets under that key. AID remains DNS-authority-rooted: the DNS owner publishes the current endpoint and current endpoint-proof key inside a TXT payload.¶
Reviewer-facing language should say "DNS-current endpoint/key" and should not imply Pkarr-style self-certifying names, key-addressed identity, or cryptographic continuity across key changes.¶
This revision changes the draft from the AID v1.2 `aid1` wire format to the AID v2 `aid2` wire format.¶