On July 23, the Internet Engineering Task Force held a Birds-of-a-Feather session called agentproto at IETF 126 in Vienna, exploring whether agent-to-agent communication protocols belong under IETF governance. The session ran as one of five BoFs at the meeting, and three of the five touched AI agents directly, but agentproto drew the most attention from people building multi-agent systems. A BoF isn't a vote and it isn't a product launch. It's an early, community-wide conversation about whether a topic is mature enough to justify a chartered Working Group. The IETF operates by rough consensus rather than ballots, and any resulting RFC is realistically years out. What's notable isn't the speed — there isn't any yet — it's the framing. Protocols like MCP and A2A, which started as vendor-driven specs from Anthropic and Google, are now being evaluated as internet infrastructure rather than platform features. That shift tracks with what's already happening inside those protocols. A2A shipped its v1.0.0 stable specification on March 12, and the release removed legacy OAuth implicit and password flows, added Device Code and PKCE support, and formalized Agent Card signature verification using JWS and JSON Canonicalization — meaning an Agent Card's identity claims can now be cryptographically checked rather than taken as self-declared metadata. MCP's maintainers published a 2026 roadmap in March acknowledging the protocol had outgrown its original scope, then moved to the Agentic AI Foundation under the Linux Foundation and expanded to handle a growing backlog of Specification Enhancement Proposals. A release candidate published May 21 was described as the largest revision since the protocol's launch, adding a stateless core and six SEPs hardening authorization around how OAuth 2.0 and OpenID Connect actually get deployed. The pattern across all of this is the same: routing and handoff mechanisms that used to be implicit — which agent talks to which, on whose authority, over what transport — are being made explicit and auditable at the protocol layer. An IETF Working Group, if agentproto gets chartered into one, would formalize that further by giving agent-to-agent communication the same standards-track treatment as email or HTTP, rather than leaving it to whichever vendor consortium ships first. That matters less for what agents can already do and more for whether a message from one agent to another, running on different infrastructure and built by different vendors, can be trusted to mean what it says without a shared corporate parent underwriting the handshake. Nothing here is finalized. The BoF outcome — whether the IETF area directors greenlight a Working Group charter — hasn't been published yet as of this writing.