
Linux Foundation hosts DNS-based AI agent identity standard
The Linux Foundation is hosting Agent Name Service, a seven-document DNS standard backed by 11 organizations for verifiable AI-agent identities, public audit logs and agent-to-agent authentication.
- By
- Sophie Lindqvist
- Filed
- Channel
- AI & Compute
- Read
- 3 min read
By Sophie Lindqvist
Eleven organizations contributed to or adopted the Agent Name Service (ANS), a seven-document open standard the Linux Foundation said on Nov. 10 it is hosting to make AI agents discoverable through the Domain Name System and give them verifiable identities. The contributors and adopters are Cisco, Cloudflare, DigiCert, DistributedApps.ai, GoDaddy, HOL, Infoblox, Radix, Salesforce, OWASP and Trustmodel.ai.
What does ANS standardize?
ANS publishes agent identities through DNS. It verifies identities with certificates. It records them in public logs for audit. It also lets agents authenticate one another. The standard combines naming, technical verification and an audit trail in one protocol layer.
The seven documents cover domain-anchored names, registration and versioning, publication in DNS, public-log auditing, monitoring of identity records for unauthorized changes, and agent-to-agent authentication within an open specification hosted by the Linux Foundation.
An agent name hangs off a domain controlled by its owner. A certificate backs the identity. DigiCert's Brian Trzupek called that arrangement "a passport for AI agents." The announcement does not specify certificate issuance, rotation or revocation rules, so the passport analogy leaves practical assurance gaps.
Who is building ANS, and when?
Linux Foundation CEO Jim Zemlin said AI agents "can't succeed without verifiable identity." Cloudflare CTO Dane Knecht said organizations "need to know who an agent represents before trusting it to act." He also said ANS uses infrastructure the internet already relies on, rather than another proprietary system.
Taken together, their claims move ANS beyond a naming experiment, but neither statement provides adoption data, implementation costs or evidence that enterprises will trust the certificates. DistributedApps.ai CEO Ken Huang said he led the initial development. ANS announced its intent to launch in June.
The announcement did not identify the year or make June a firm production date. GitHub hosts the specification and its code. Neither fact proves broad implementation or production readiness.
The Linux Foundation's AGNTCY Agent Directory Service can verify ANS identities. An implementation called ANS Finder passes conformance tests for the Agentic Resource Discovery specification. The announcement gives no test totals, performance data or production-user figures for either implementation.
How does ANS differ from MCP?
ANS addresses a different layer from Anthropic's Model Context Protocol. Anthropic donated MCP to the Linux Foundation, which already hosts it. MCP connects agents to tools and data. ANS instead addresses who an agent is and whom it represents. The announcement presents the two specifications as complementary, not as substitutes for identity controls.
Other development continues in the same field, including Oracle's Open Agent Specification. Infoblox said domains can serve as the naming and discovery layer "for agents and MCP servers." That choice puts DNS operators, registrars and certificate authorities near the commercial control point for agent identity, where proprietary trust services now compete with open naming systems.
What is driving enterprise interest?
Corporate caution supplies part of the commercial rationale. CIO Dive reported that some financial institutions have begun closing their systems to outside agents. The report did not name those institutions or quantify the restrictions. United Airlines barred AI agents from booking its flights in the week reported alongside ANS, while OpenAI has started letting businesses run verified agents in ChatGPT.
These actions do not prove that enterprises have adopted ANS. They show different control choices. One airline blocked agent bookings. A chat platform enabled verified business agents. Neither example measures demand for DNS-based agent identity, and neither connects directly to ANS deployments.
Can ANS turn controls into trust?
The announcement offers no adoption count, customer list, pricing model or deployment schedule beyond the intended June launch. Its strongest claim is architectural: an open identity layer can reuse DNS, certificates and public logs already familiar to internet operators.
The competitive test will be whether ANS Finder, the AGNTCY directory and the planned June release turn that interoperability into enterprise trust without recreating a proprietary control point.
Source: mlq.ai
More from Sophie Lindqvist
Show full bio
News editor covering money, funding and market shifts around Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
919 articles
Related articles
nebius-beats-revenue-estimates-then-raises-capex-to-27-billion
Nebius Beats Revenue Estimates, then Raises Capex to $27 Billion
cerebras-q3-test-25-4-billion-backlog-meets-margin-risk
Cerebras Q3 Test: $25.4 Billion Backlog Meets Margin Risk
cisco-posts-9-3-billion-in-ai-orders-ahead-of-q1-fy2027-earnings
Cisco Posts $9.3 Billion in AI Orders Ahead of Q1 FY2027 Earnings
legence-q3-revenue-jumps-85-as-data-center-backlog-reaches-5-9b
Legence Q3 revenue jumps 85% as data-center backlog reaches $5.9B



