Forty-nine percent of cloud-native teams are running AI agents in production right now. Most of those agents have no verifiable identity — and 68% of organizations can’t distinguish agent activity from human activity at all. That gap is the reason the Linux Foundation launched Agent Name Service (ANS) on November 10, 2026 at KubeCon + CloudNativeCon North America. It’s an open standard that gives AI agents DNS-anchored, verifiable identities, and eleven organizations are already behind it: Cisco, Cloudflare, Salesforce, GoDaddy, DigiCert, Infoblox, and OWASP among them.
The Problem: You Can’t Govern What You Can’t Name
AI agents are transacting across global systems, calling APIs, reading files, and triggering workflows. But ask most organizations which agent did what, and they can’t answer. A Cloud Security Alliance survey found 68% of organizations cannot reliably distinguish an AI agent’s activity from a person’s. DataDome found 80% of agents don’t prove who they are — and 80% of sites don’t bother checking. Worse, 28% of organizations have already experienced an AI agent taking an unapproved action.
The industry has been treating agent identity as a problem to solve ‘later.’ ANS is the argument that later is now.
What ANS Does
ANS gives every AI agent a verifiable, versioned name anchored to a DNS domain the operator controls — something like ans://v1.0.0.my-agent.example.com. To get that name, the agent’s domain goes through the same ACME certificate verification process that websites use to get TLS certificates. No new registrar required. No gatekeeper.
Once registered, every lifecycle event — creation, versioning, deprecation — gets sealed into an append-only Transparency Log built on SCITT receipts. You can audit it. You can verify it offline using the ans-verify tool. And discovery is handled through DNS-AID, a companion standard that lets agents publish their endpoints as plain DNS records so other agents can find them.
The full spec spans seven layers: identity anchoring, registration, versioned naming, DNS publication, transparency logging, integrity monitoring, and agent-to-agent authentication. The spec and open source code are on GitHub.
Why DNS Is the Right Call Here
The competing approaches — decentralized identifiers (DIDs) and in-cluster standards like WIMSE/SPIFFE — each have legitimate use cases. SPIFFE excels at ephemeral, cryptographic workload identity inside a cluster. DIDs avoid any single identity provider. But neither solves the cross-organization, human-readable, discoverable naming problem.
ANS does. It uses the trust model the entire internet already relies on. Every organization with a domain name can register agents under it immediately, with the same infrastructure they already operate. There’s no proprietary namespace and no central authority that can be captured or monopolized. Brian Trzupek, the DigiCert VP who led the initial spec development, put it plainly: ‘By hosting ANS under neutral governance, the Linux Foundation gives the whole ecosystem a trusted home for agent registration, discovery, and security governance.’
The Ecosystem Already Building Around It
ANS doesn’t exist in isolation. Cisco’s open-source AGNTCY project — the ‘Internet of Agents’ directory now also under the Linux Foundation — integrates ANS directly. ANS handles trusted identity; AGNTCY handles discovery. Together, they give multi-agent systems something they’ve been missing: a way to find another agent and verify that it is who it claims to be.
This announcement came at KubeCon NA 2026, where the CNCF reported 49% of cloud-native developers have agentic systems in production — and where Google announced Agent Substrate, Mirantis launched a governed agent platform, and Red Hat shipped OpenShift 5 with built-in agent tooling. Agent governance is not a future problem. It’s this quarter’s problem.
What Developers Should Watch
ANS is infrastructure layer zero for the agent ecosystem — the layer that has to exist before governance, auditing, and cross-org agent federation can be built. Expect integrations with Model Context Protocol (MCP) and Google’s Agent-to-Agent (A2A) standard over the next 12 months as those protocols need to reference agent identities.
If you’re building multi-agent systems, the spec and reference implementation are live at github.com/agentnameservice. The Linux Foundation announcement includes the full list of supporting organizations and the roadmap. The IETF draft (draft-narajala-ans-00) is in progress for formal standardization.
Agents need passports. ANS is the first credible attempt to issue them.
