Discovery · DNS
IETF Approves Charter for a Dedicated Agent Discovery Working Group
On October 8, 2026, the IETF's Internet Engineering Steering Group approved the charter for a new working group, agentproto (Agent Communication Protocols), giving agent discovery a formal home inside the standards body that runs the internet's core protocols. As of the following day, the group had no working documents attached yet — the charter exists, the drafts don't, at least not officially.
What the charter actually commits to is notable for what it avoids promising: rather than inventing a new discovery mechanism from scratch, the proposed charter covers the discovery of AI agents and says it will use existing IETF protocols where possible. That's a meaningful constraint. A lot of the agent-discovery conversation over the past year has assumed a new registry layer is needed — Google's Agent Registry, HOL's Universal Agent Registry, various commercial catalogs. The IETF's instinct here is the opposite: reuse DNS, reuse existing lookup and naming infrastructure, and treat agent discovery as an extension of problems the internet has already solved rather than a greenfield design exercise.
The same standards tracker that reported the agentproto charter also flagged a second, related effort moving through IETF process: a BOF request on agent delegation and interaction traceability, approved the same day, October 8. That request touches on correlating delegation chains and authorization state across domains — the problem of knowing which agent acted on whose authority when a task passes through several hops. It's a separate track from agentproto, still at the BOF (Birds of a Feather) stage rather than a chartered working group, and an earlier version of the same proposal had been declined back in July.
Neither effort has shipped a document yet. That's worth sitting with rather than glossing over: a charter approval is a commitment to do the work, not the work itself. The tracker notes that as of October 9, the datatracker API listed zero working-group documents for agentproto. Calls for adoption and meeting agendas are not the same thing as an adopted standard, and the gap between those two states is where most agent protocol proposals have stalled over the past two years — A2A, MCP extensions, and various authorization drafts have all moved through individual-draft purgatory before (or instead of) reaching WG consensus.
What makes this charter approval worth noting now is less the content and more the venue. Agent discovery has mostly been fought over in vendor blog posts and framework repos. Having the IETF formally stand up a working group for it means the question of how agents find each other is now, at least nominally, subject to the same multi-stakeholder process that produced DNS, HTTP, and TLS — slower, but harder to capture by any single vendor.