GoDaddy announced today its intent to submit to the Linux Foundation an implementation of an open specification that provides the foundation for creating a global naming service for artificial intelligence (AI) agents.
Announced at the Open Source Summit North America conference, the AI Agent Naming (ANS) Service is an extension of the Domain Name System (DNS) platform previously authored by GoDaddy, Cisco and OWASP that defines a specification for a secure registry for AI agents that is being advanced under the auspices of the Internet Engineering Task Force (IETF).
ANS relies on Public Key Infrastructure (PKI) and SSL certificates to validate an AI agent. These standardized naming structures allow agents to locate, evaluate, and negotiate with other agents based on specific functions, permissions, and protocols.
Specifically, a draft of that specification specifies how agents can publish discoverable metadata using existing DNS record types, including RFC 9460 service bindings (SVCB), DNS-SD service discovery, domain name system security extensions (DNSSEC), and DNS-based authentication of named entities (DANE).
Oliver George, vice president of strategic partnerships for GoDaddy, told conference attendees that this approach will make it possible to ensure that there is a clear owner of an AI agent that has a set of known permissions. “Without this infrastructure, AI agents are anonymous,” he says. “They are opaque.”
The primary reason GoDaddy is pressing a case for using DNS is that it has already proven it works at scale at sub-100-milliseconds, notes George.
Additionally, GoDaddy is working with Infoblox to advance DNS for AI Discovery (DNS-AID), an open framework for agent discovery built on existing DNS infrastructure.
It’s not clear how interoperability between AI agents will be achieved and maintained. There are already multiple protocols and interfaces being proposed, including:
A Model Context Protocol (MCP) through which servers advertise their tools in a way that makes it simpler for AI agents to discover them automatically.
An Agent2Agent (A2A) protocol that standardizes how agents discover and communicate with each other.
A Universal Commerce Protocol (UCP) that standardizes the shopping lifecycle via request and response schemas that remain consistent across any underlying transport.
An Agent Payments Protocol (AP2) that provides non-repudiatable proof of intent and enforces configurable guardrails on every transaction.
An Agent-to-User Interface Protocol (A2UI) that uses JSON, enabling an AI agent to dynamically compose layouts from a fixed catalog.
The degree to which organizations will need robust implementations of all these protocols to operationalize agentic AI workflows remains to be seen, but it’s clear much work remains to be done. The challenge, as always, is that organizations are often reluctant to build applications if de facto standards have not been clearly established.
At the same time, pressure is mounting to adopt agentic AI technologies as quickly as possible. Organizations are looking to reduce costs by relying on AI agents to perform rote tasks, while also experimenting with new use cases that expand the scope and reach of the digital services they provide.
The larger issue is mastering the nuances of a set of rapidly emerging technologies that are still subject to significant change.
