The Linux Foundation has taken over Agent Name Service, an open specification that gives AI agents domain-anchored names and a verifiable identity — and draws a firm line between proving who an agent is and deciding what it may do.
What happened
On 10 November, at KubeCon + CloudNativeCon North America in Salt Lake City, the Linux Foundation announced that Agent Name Service (ANS) had launched as a foundation project. ANS is an open standard built on DNS infrastructure for naming, discovering and authenticating AI agents.
The specification is published as seven documents covering domain-anchored naming, registration and versioning, DNS publication, public log auditing, integrity monitoring, and agent-to-agent authentication. A registration is identified by a version and a domain, written in the form ans://v1.0.0.payments.example.com. A new version is a new registration.
The lifecycle works like this. An operator submits an agent host, version, endpoints and certificate material to a Registration Authority, which verifies domain control and the required proofs. The operator publishes the DNS records the authority returns. The authority signs the lifecycle event and a Transparency Log seals it before activation succeeds, issuing a badge or a SCITT receipt that binds the registration to its authorised certificates.
Verification rests on three separate proofs. Identity asks whether the certificate is bound to this agent in the transparency log. Current status asks whether the registration is still valid, answered by an online badge or a short-lived signed status token with bounded freshness. Possession asks whether the peer actually holds the private key — a public badge or receipt alone cannot authenticate a caller. ANS-6 defines two caller authentication methods with equal standing: mutual TLS presenting the ANS Identity Certificate, or application-layer DPoP, which passes through gateways that terminate TLS. DNS itself is used for discovery, through the ANS_DNSAID profile by default in the current V2 model, an optional ANS_TXT record, a required _ans-badge record pointing at the transparency log, and an optional DANE TLSA record.
Eleven organisations have contributed to or adopted the standard since the June intent to launch: Cisco, Cloudflare, DigiCert, DistributedApps.ai, GoDaddy, HOL, Infoblox, Radix, Salesforce, OWASP and Trustmodel.ai. The AGNTCY Agent Directory Service, another Linux Foundation project, supports verification of ANS identities, and ANS Finder implements the Agentic Resource Discovery (ARD) specification and passes its registry conformance tests.
“The specification is now formalized and will be governed by the project’s technical steering committee,” the foundation said. The project site is more cautious: it labels the published specifications as drafts under active development.
Two boundaries in the specification matter as much as the mechanism. ANS verifies a domain-anchored identity; optional “Verified Identities” can add a separately proven identity relationship where a Registration Authority supports them, and sealed metadata digests bind published metadata to a registration without attesting the code running on a remote machine or guaranteeing an agent’s behaviour. And, explicitly: “ANS authentication does not authorize an action or delegate a user’s permissions. Trust scoring is a separate project, not an ANS authentication tier.”
Why it matters for KYA
Know Your Agent needs a naming layer before it can have a verification layer. You cannot check the authority of an agent you cannot name, and you cannot audit an interaction with an agent whose identifier is whatever string it put in a header. ANS is an attempt to make that name resolvable, domain-anchored and publicly auditable using infrastructure that already exists — which is why the supporters list reads like the DNS, CDN, certificate and identity vendors rather than the model vendors.
For practitioners, the interesting part is the shape of the evidence. The three proofs are separable on purpose. A certificate in a log proves binding, a status token proves the registration has not lapsed, and a key-possession check proves the caller is the party named in the log. Systems that check only the first — a valid certificate, cached at some point in the past — will accept agents whose registration has since been revoked or changed. That is the practical failure mode ANS is designed to make visible, and it is why the specification requires the status check alongside the identity check.
The second point is the boundary. ANS stops at authentication, and says so. It does not carry the user’s delegated permissions, it does not decide whether the requested action is within scope, and it does not score trust. Those remain the relying party’s problem, to be solved with delegation records, scopes or payment mandates and enforced in policy at the point of action. An ANS identity is a prerequisite for KYA, not a substitute for it: it answers “which agent is this and who operates it”, not “what is it allowed to do for which principal, and what happens when that authority is withdrawn”.
The domain anchor is also worth being clear-eyed about. A domain proves control of a name, not accountability for an agent’s actions. It gives a relying party a party to talk to and a record to audit — a meaningful improvement over self-declared agent names — but the accountable principal, the scope of authority and the audit trail of actions still have to be established by the systems the agent touches.
What to watch
- Whether the seven documents converge out of draft, and how the project handles revocation and status freshness at scale.
- Adoption beyond the founding supporters: whether major agent platforms and payment schemes accept ANS identities at the edge, or keep proprietary registries.
- Interoperability with adjacent work — Agentic Resource Discovery, AGNTCY, Web Bot Auth and the OAuth-based delegation drafts — since a name that only one ecosystem can resolve solves little.
Sources
- The Linux Foundation, “Linux Foundation Welcomes Agent Name Service to Bring Verifiable Identity and Discovery to AI Agents”, 10 November 2026 — linuxfoundation.org
- Agent Name Service, “How ANS works” — ansinfo.ai
- Agent Name Service, ANS-6 agent authentication specification — github.com/agentnameservice
- Agent Name Service, ANS-3 DNS publication specification — github.com/agentnameservice
Know Your Agent (KYA) explains agent identity, verification and accountability. This is an explainer, not legal or compliance advice — see our Sources & methodology. New to KYA? Start with What is Know Your Agent? and the glossary.