AI agents keep multiplying. Yet they struggle to locate each other without custom lists or fragile links. The Linux Foundation now offers a fix drawn from infrastructure that has worked since the 1980s.
On May 27, 2026, the organization announced the DNS-AID project. The effort, first shaped inside Infoblox, turns the Domain Name System into a shared directory for autonomous software. No fresh databases. No single company in control. Just DNS records that agents query the same way browsers resolve names.
Current setups leave much to be desired. Developers hard-code endpoints. They maintain private catalogs. Or they accept walled gardens run by big platforms. Those choices create friction. They raise costs. They limit what agents can achieve when they need to cooperate across organizations.
DNS already knows how to scale discovery.
Think about email. Servers publish SPF, DKIM and DMARC records in DNS so recipients can verify senders without asking anyone first. Service Binding records, standardized in RFC 9460, let clients learn connection parameters before they even open a socket. DNS-AID builds on those patterns.
Agents publish metadata at predictable names such as _{agent-name}._{protocol}._agents.{domain}. The proposal also points to a starting index at _index._agents.{domain}, a spot any visitor can check first. Supporting records include SVCB with TXT as fallback, plus DNSSEC for validation and DANE for cryptographic proof of the endpoint. The Register detailed this mechanism in its May 28 report. (The Register)
Ingmar Van Glabbeek, one of the project maintainers, captured the shift. “Current approaches to agent connectivity are fragmented and often rely on fragile, hardcoded configurations. With DNS-AID, we are moving toward a ‘web-native’ model for AI. By using the existing DNS hierarchy, we enable developers to publish and discover agents with the same reliability and ubiquity that we’ve used to navigate the internet for decades.”
The Linux Foundation hosts the work to keep it neutral. Initial backers read like a who’s who of internet infrastructure: Cloudflare, GoDaddy, CSC, Equinix, Indeed, Internet Systems Consortium, WWT and others. Jim Zemlin, CEO of the Linux Foundation, framed the stakes. “AI agents are quickly becoming the connective tissue of the modern internet, but without secure, open discovery infrastructure, that connectivity becomes a liability. DNS-AID helps anchor agent discovery in the DNS infrastructure the internet already trusts.”
Dane Knecht, CTO at Cloudflare, added perspective. “The Internet already solved the discovery problem decades ago with DNS – it’s fast, it scales globally, and every network on earth understands it. By extending this proven architecture to the agentic web, DNS-AID provides the foundational routing layer that autonomous systems need to operate safely and efficiently.”
Infoblox and GoDaddy had signaled their interest even earlier. On May 14 they jointly backed two linked standards. DNS-AID handles discovery and capability advertising. A companion effort called Agent Name Service, or ANS, focuses on identity and verification using domain names operators already own plus public-key infrastructure. The GoDaddy release spelled out the pairing. (GoDaddy press release)
Wei Chen, Chief Legal Officer at Infoblox, drew a historical parallel. DNS replaced the phone book. It became the gold standard for digital trust. Extending it to agents simply continues that tradition.
InfoWorld noted the contrast with commercial offerings. Several proprietary agent registries already exist. AWS offers one inside Bedrock. Others push their own directories. The Linux Foundation argues the open path avoids vendor lock-in and duplicated effort. (InfoWorld)
Implementation arrives ready to test. The project ships a Python SDK, command-line tools and an MCP server reference. Model Context Protocol servers can advertise themselves alongside agents. Support already appears in major DNS services including AWS Route 53, Azure DNS, Cloudflare, Google Cloud DNS and Infoblox systems. Developers can spin up a local BIND9 container for experiments.
Security sits at the center. DNSSEC signatures vouch for data integrity. DANE records tie certificates to names. Optional JSON Web Signatures add another layer. Agents can check provenance before they exchange messages or trigger actions. That matters when agents start negotiating contracts, moving money or sharing sensitive data.
Adoption will not happen overnight. Details still need refinement. The project lives on GitHub at github.com/dns-aid and invites contributions. IETF drafts advance the technical specifications. Yet the core idea feels familiar to network engineers. Reuse what works. Avoid reinventing trust at internet scale.
McKinsey has floated large numbers for agent-to-agent commerce. The consultancy sees trillions of dollars in potential economic value. History suggests such forecasts carry uncertainty. Still, the direction looks clear. Autonomous systems will interact more. They will need a common directory that does not depend on any one platform.
Slashdot first surfaced the story to a wide technical audience on May 31. (Slashdot) Discussion on X echoed the coverage, with links spreading quickly among developers and security professionals.
The project does not promise miracles. It offers a practical extension of infrastructure already deployed everywhere. Agents gain a way to announce who they are, what they do and how to reach them. Humans keep control through the domains they already register and the DNS zones they manage.
That simplicity carries power. Enterprises can publish internal agents without exposing them to public search. SaaS companies can let customer agents find specialized helpers. Developers can query for capabilities instead of maintaining endless integration tables. The directory grows with the web itself.
Challenges remain. Governance questions will surface as the community expands. Privacy considerations arise when agents advertise capabilities. Interoperability tests must prove the records behave consistently across resolvers. Yet these look like engineering tasks, not architectural dead ends.
DNS survived every internet transformation so far. It absorbed email, the web, voice over IP and cloud workloads. Adding AI agents to its responsibilities fits the pattern. The directory was always there. The agents just learned how to read it.
WebProNews is an iEntry Publication