---
title: Agent Name Service (ANS)
type: vocabulary
url: "https://www.envisioning.com/vocab/agent-name-service"
summary: A DNS-based protocol that gives AI agents verifiable identities, owners, and permissions.
year: 2026
generality: 0.55
---

# Agent Name Service (ANS)

A DNS-based protocol that gives AI agents verifiable identities, owners, and permissions.
An Agent Name Service is a federated identity and discovery layer for AI agents that uses the Domain Name System as its underlying registry. Where DNS translates human-readable hostnames into IP addresses, ANS translates an agent's name into a signed manifest that declares who owns the agent, what it is allowed to do, and how its code can be verified. By publishing records under a domain the operator already controls, an organization can register an agent at a name like agents.example.com without relying on a new centralized authority.

ANS works by combining two complementary identity markers. A Decentralized Identifier (DID) lets the agent prove cryptographically that its runtime has not been tampered with since it was signed. A Legal Entity Identifier (LEI) ties that cryptographic identity to a real-world corporate owner, so a counterparty system can confirm not just which code is calling but which accountable entity stands behind it. When one agent queries another, the request resolves through DNS to the manifest, the manifest returns the DID and LEI, and the calling system can decide whether to trust the agent before granting access.

The main trade-off of building on DNS is that the protocol inherits DNS's strengths and weaknesses. Adoption is cheap because every enterprise already runs DNS, and there is no proprietary registry to negotiate with, which keeps the agentic web open rather than walled-garden. The cost is that DNS itself is a 1980s protocol vulnerable to spoofing and hijacking, and propagation delays mean an agent whose manifest changes can be temporarily locked out of the systems it depends on. Most deployments treat ANS as the public phonebook layer that pairs with private controls like Identity and Access Management and AI gateways rather than as a standalone trust anchor.

Several open questions remain. ANS is one of multiple competing standards in what analysts call the standards discovery phase, alongside the Model Context Protocol for tool connectivity, the Cisco-led AGNTCY project for messaging, and the DNS-AID project for capability advertising. It is not yet clear whether identity, tool access, and observability will converge into one protocol or remain separate layers, nor how regulators will treat cryptographic agent identities when agents cause real-world harm. The framework is also untested against sophisticated DNS-level attacks aimed at redirecting traffic to rogue manifests, and the operational burden of keeping manifests in sync with fast-moving agent code is still being measured.

---
Source: Envisioning — Technology Research Institute (https://www.envisioning.com/vocab/agent-name-service)