Agent Name Service (ANS), an open standard that uses DNS and existing certificates to name, register and verify AI agents, has launched as a Linux Foundation project, the foundation announced on November 10 at KubeCon North America.
What happened
The specification runs to seven documents covering domain-anchored naming, registration and versioning, DNS publication, public log auditing, integrity monitoring and agent-to-agent authentication. A technical steering committee now governs it. The foundation lists 11 organizations that have contributed to or adopted ANS, including Cisco, Cloudflare, DigiCert, GoDaddy, Infoblox, Salesforce and OWASP. AGNTCY’s Agent Directory Service already verifies ANS identities, and ANS Finder passes the registry conformance tests of the Agentic Resource Discovery specification.
Why it matters
ANS covers one question: whether an agent arriving from outside the company is the agent it claims to be. DigiCert’s Brian Trzupek called it “a passport for AI agents” in the announcement. A passport says who an agent is. It says nothing about what that agent may do for a given user, or how fast that permission ends.
That second half sits with the identity provider. In an AWS test of Okta cross-app access with Bedrock AgentCore, Okta refused an agent’s new requests within about 1 second of an administrator removing a user, and tokens already issued expired within 5 minutes. Google’s Gemini Agent shipped with its own agent identities inside its platform.
For security architects and procurement teams, the practical step is to ask agent vendors and partners two things separately: whether their agents publish a verifiable ANS identity, and how their delegated permissions are scoped and revoked. A yes to the first does not answer the second.
Source: Linux Foundation, Linux Foundation Welcomes Agent Name Service. Additional source: AWS Machine Learning Blog, Govern AI agent access with Okta cross-app access and Amazon Bedrock AgentCore.
