The rapid proliferation of AI agents across the internet is creating a new challenge: how do these autonomous programs find each other to communicate and collaborate? The Linux Foundation believes the answer lies in the same distributed system that already powers the web — the Domain Name System (DNS). The foundation has announced DNS-AID, a proposal for a standardized way for AI agents to discover and verify one another using DNS without requiring any new infrastructure.
DNS-AID stands for DNS for AI Discovery, and it is designed to work with the existing DNS protocol. The core idea is that domain owners can create a well-known address — _index._agents.{domain} — that serves as a starting point for agents looking for other agents within that domain. This approach leverages the scalability, resilience, and ubiquity of DNS, which has been the backbone of internet naming for decades.
The Growing Need for Agent Discovery
AI agents are becoming increasingly common in areas such as customer service, data analysis, automation, and even creative work. These agents often need to interact with each other to complete complex tasks, such as coordinating supply chains, managing smart buildings, or processing real-time data streams. Without a standardized discovery mechanism, developers must rely on proprietary registries, hard-coded endpoints, or centralized directories — all of which introduce vendor lock-in, single points of failure, and security risks.
DNS-AID aims to solve this by providing a global, open, and vendor-neutral directory that any agent can query. The Linux Foundation notes that the internet already trusts DNS for critical operations like website resolution and email routing, so extending it to AI agents is a natural evolution. The proposal ensures that agent discovery remains compatible with the underlying internet infrastructure, making it easier for agents from different ecosystems to interact.
How DNS-AID Works
According to the draft specification, an AI agent looking for other agents would first query the DNS for _index._agents.{domain} within the target domain. The DNS response would contain records pointing to the available agents, along with metadata such as their capabilities, protocols, and verification keys. This allows agents to not only find each other but also verify their identities and establish secure communications.
The system is designed to work with the Model Context Protocol (MCP), a protocol that standardizes how AI models and agents exchange context. By integrating with MCP, DNS-AID enables agents to discover MCP servers and other agent endpoints seamlessly. The Linux Foundation emphasizes that no new infrastructure is required — only a DNS server that can handle the new record types, and agents that can parse them.
Industry Support and Initial Development
DNS-AID was initially developed by staff at Infoblox, a company specializing in DNS and network services. The latest internet draft includes contributions from engineers at Deutsche Telekom and Amazon, indicating broad industry interest. The Linux Foundation, which hosts numerous open-source projects including the Kubernetes ecosystem and the Cloud Native Computing Foundation, will oversee the project's governance to ensure it remains vendor-neutral.
Jim Zemlin, CEO of the Linux Foundation, highlighted the importance of open infrastructure: “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 that the internet already trusts.”
Background: The Evolution of DNS
DNS was originally developed in the 1980s to solve the problem of translating human-readable domain names into machine-readable IP addresses. Over the years, it has been extended to support a wide variety of applications, including email routing (MX records), service discovery (SRV records), and security (DNSSEC). The DNS-AID proposal continues this tradition by adding new record types and query patterns specifically for AI agent discovery.
The open nature of DNS has been critical to its success. Unlike proprietary directories that can be altered or shut down by a single organization, DNS is maintained by a global community of operators and governed by standards bodies such as the Internet Engineering Task Force (IETF). This decentralized model ensures resilience against attacks and censorship, and it scales to billions of queries per day.
Challenges and Security Considerations
While DNS-AID leverages existing infrastructure, it also introduces new security challenges. Malicious actors could attempt to register fake agent entries or intercept discovery queries. The proposal addresses this by integrating with DNSSEC (DNS Security Extensions), which provides cryptographic verification of DNS responses. Agents can verify that the records they receive are authentic and have not been tampered with.
Another challenge is scalability. As the number of AI agents grows into the billions, DNS servers must handle increased query volumes. However, the distributed nature of DNS, combined with caching and load balancing, makes it well-suited for this task. The Linux Foundation is working with DNS operators to ensure that the new record types do not overwhelm existing infrastructure.
Comparison with Other Approaches
Several companies have launched proprietary agent registries, such as Google’s Agent Registry and various blockchain-based directories. While these solutions offer advanced features, they suffer from interoperability issues and vendor lock-in. DNS-AID takes a different approach by building on an existing open standard that is already universally deployed. This reduces the barrier to entry for developers and allows agents from different vendors to communicate without needing to agree on a third-party platform.
Additionally, DNS-AID is protocol-agnostic. It does not require agents to use a specific messaging format or AI model. As long as an agent can perform a DNS query and parse the response, it can participate in the discovery ecosystem. This flexibility is crucial in a rapidly evolving field where new protocols and standards emerge regularly.
Integration with MCP and Other Protocols
The Model Context Protocol (MCP) is an emerging standard for sharing context between AI models and agents. DNS-AID complements MCP by providing a discovery layer. An MCP server can register its endpoint in DNS-AID, and agents can find it without prior knowledge. This combination simplifies the process of building multi-agent systems and allows for dynamic reconfiguration as agents come online or go offline.
Other protocols, such as the Agent Communication Protocol (ACP) from the AI Alliance, could also benefit from DNS-AID. The Linux Foundation hopes that DNS-AID will become a foundational layer that supports multiple upper-level protocols, similar to how DNS underpins HTTP, SMTP, and many other internet protocols.
The Role of the Linux Foundation
The Linux Foundation has a track record of shepherding open-source projects that become industry standards. By hosting DNS-AID, the foundation ensures that the project remains open and governed by a community of contributors rather than a single vendor. This is particularly important for AI infrastructure, where trust and transparency are paramount.
The foundation is now inviting contributions to the DNS-AID project, including feedback on the draft specification, implementations in various programming languages, and test deployments. The goal is to produce a stable standard that can be adopted by DNS server software providers and AI agent frameworks.
Implications for Developers and Enterprises
For developers building AI agents, DNS-AID simplifies the discovery process and reduces reliance on proprietary APIs. An agent can be configured with a list of target domains and automatically find available agents within those domains. This enables more dynamic and resilient architectures where agents can be added or removed without updating configuration files.
Enterprises that deploy multiple AI agents for different business functions can use DNS-AID to create a unified directory within their own domain. For example, a company could have agents for customer support, inventory management, and fraud detection, each registering under _index._agents.company.com. External partners could also be granted access to specific subdomains, enabling secure inter-company communication.
The scalability of DNS means that even large enterprises with thousands of agents can rely on the same infrastructure that already handles their website traffic. This reduces operational complexity and cost.
Future Directions
While DNS-AID is still in the early stages, the Linux Foundation has outlined a roadmap that includes support for advanced features such as agent capability discovery, real-time status updates, and reputation systems. Future drafts may also incorporate mechanisms for anonymous or pseudonymous agents, which could be important for privacy-sensitive applications.
The project team is also exploring integrations with blockchain for immutable agent identity records, though they emphasize that DNS-AID itself remains a DNS-native solution. The modular design allows for innovation at the application layer without changing the core DNS protocol.
As the internet of agents grows, the need for robust discovery will only increase. DNS-AID offers a pragmatic and proven approach that leverages the best of what the internet already has to offer. By keeping the infrastructure open and vendor-neutral, the Linux Foundation hopes to prevent the fragmentation that has plagued other areas of technology.
Developers interested in contributing can find the draft specification on the Linux Foundation’s website and participate in the working group that will refine the standard. With input from telecom providers, cloud giants, and DNS experts, DNS-AID has the potential to become the standard way for AI agents to find each other on the open internet.
Source: InfoWorld News