Name: Facilitating AI Agent-to-Agent Discovery Through Existing Internet Protocols

Description

The rapid proliferation of AI agents necessitates a robust, dynamic and scalable
mechanism at the networking layer to enable secure and autonomous discovery,
authentication and interaction. The purpose of this BOF is to explore leveraging and
augmenting the existing Domain Name System (DNS) protocol to achieve these goals due
to DNS’s ubiquity and scalability.

The goal of this BOF is to brainstorm on how to utilize today’s decentralized internet body
(certificates, ASNs, registrars/DNS, etc.) to facilitate agent-to-agent discovery while
maintaining operator sovereignty, and addressing the need for authentication, privacy,
security, scalability and low latency.

A few research proposals explored the need for a new protocol that includes billing,
authentication, logging, and other mechanisms which then go on to propose DNS but in a
less performant manner. Before chartering work on new protocols, it is prudent to explore
a path utilizing existing DNS technologies such as Service Binding records, DANE,
DOT/H/Q, DCV (DNSOP almost-published standard), DNSSEC, special-use domain
names, gTLDs, etc.. Exploring the mechanisms of how DNS could facilitate agent-toagent
discovery, instead of developing an entirely new protocol and standard, allows for quicker
development to focus efforts on the application transaction, and not the establishment of
a new session.

In this BOF, we will assess the current capabilities and limitations of DNS protocols and
the needed enhancements for AI agent-specific exchanges. The outcome should be a
clear understanding of where the gaps lie and whether they can be addressed through
augmenting the existing protocols.

This BOF is designed to spark discussion across the IETF community. We aim to gather
feedback, share experiences, and address critical questions around the scope of this
work.

Required Details

• Status: non-WG forming
• Responsible AD: internet area, Eric Vyncke
• BOF proponents: Nic Williams (nwillau900@gmail.com), Jim Mozley
  (jmozley@infoblox.comr)
• Number of people expected to attend: 100
• Length of session (1 or usually 2 hours): 2 hours
• Conflicts (whole Areas and/or WGs)
• Chair Conflicts: TBD
• Technology Overlap: TLS, DNS
• Key Participant Conflict: employees of Infoblox, a DNS provider.

Information for IAB/IESG

• Any protocols or practices that already exist in this space:
  DNS
• Which (if any) modifications to existing protocols or practices are required:
  A special use domain name [RFC 6761] is to be reserved for AI agents. There are more
  potential outcomes.
• Which (if any) entirely new protocols or practices are required:
  None proposed at this time.
• Open source projects (if any) implementing this work:
  https://github.com/supervaize/supervaizer
  https://github.com/Jordiag/semantic-kernel-agent-a2a-protocol-example
  (!!!) https://github.com/agentcommunity/agent-interface-discovery
  https://github.com/allenday/a2a-registry

Agenda

• Introduction and background
• Problem statement: need for secure and scalable agent-to-agent discovery
• Discovery: unique needs of agent-to-agent discovery and communication
• Review of DNS capabilities and limitations
• Charter/scoping discussion
• BOF Q&As

Links to the mailing list, draft charter if any (for WG-forming BoF), relevant InternetDrafts, etc.

• Mailing List: https://www.ietf.org/mailman/listinfo/agent2agent
• Draft charter: N/A
• Relevant Internet-Drafts:
• Use Cases
• Solutions
  Draft will be published end of September and this link will be updated.