Summary

  • The IESG approved Agentproto's charter on 8 October 2026. The group aims to adopt a use-cases draft in December, before taking on architecture and protocol drafts in 2027.
  • A use-cases proposal updated on 15 October remains an individual Internet-Draft. Its authors have revised a document; the working group has not thereby chosen it.

An AI agent can produce an answer and still leave the next system with a basic question: which conversation does this belong to?

That problem gets harder when work moves between services, switches from speech to text or lasts longer than a single connection. Remembering the subject of a discussion is one thing. Letting several independent systems recognise the same discussion, and know whether it is still active, is another.

The IETF's Agentproto working group has been given the second job. Its approved charter calls for a common way to manage these exchanges and an architecture showing how the pieces fit together. Here, the context of a conversation means identifiers and lifecycle state. It does not mean handing over a model's memory, retrieved documents or prompts.

This is a deliberately limited ambition. Agentproto is not designing a new transport or replacing application-layer agent protocols. Existing IETF protocols are to carry the context. The group is also leaving model implementation, reasoning and planning, user interfaces and audit-data structures outside its remit. It is trying to keep a conversation connected, not prescribe what an AI should think or say.

A document is ready to read. A decision is still to come

The first test of progress is unusually easy to misread. The group's timetable gives December 2026 as the target for adopting an informational use-cases draft. Yet its related-documents list already includes a proposal called Agentic AI Use Cases and Requirements. Revision -04 is dated 15 October.

The apparent head start disappears when the document's own status is opened. It is an active individual Internet-Draft, with no RFC stream assigned. The Datatracker page explains that an individual submission is not IETF endorsement. Posting a document and choosing it as a working group's starting text are separate actions.

One detail needs care. The page's intended-RFC-status field is blank, but the draft itself says that its intended status is Informational. That is an intention stated by its authors, not evidence of adoption. The blank field does not erase the intention; the intention does not turn the proposal into an agreed deliverable.

The draft offers material worth discussing. Its scenarios include agents sharing work, a delegated task that runs long enough to need an authorisation checkpoint, and tools or data reached through intermediaries. These are proposed use cases and requirements, not evidence that the working group has accepted them or that implementations are already in service.

For a long-running task, an identifier alone cannot settle every question. Systems may need to distinguish an agent's identity from the user's and recognise when delegated access has been revoked. The charter identifies those concerns while keeping its focus on the protocol work. It does not promise that standardising conversation management will resolve every application-level security decision.

What October's approval actually bought

The charter history records IESG approval on 8 October. That gave the group its mandate. It did not approve this particular use-cases proposal, much less a finished protocol.

After December, the published plan calls for adoption of an informational architecture draft in February 2027 and one or more standards-track protocol drafts in March 2027. The target for submitting protocol drafts to the IESG as proposed standards is March 2028. The history also records the move to wording that allows multiple drafts at the later protocol stages.

Those dates describe work to be done. An updated file is useful evidence that authors are working, but cannot stand in for the group's choice of text. The next consequential entry will identify which use-cases document the group adopts, and which version the decision covers. If the timetable changes first, that will matter too.

Agentproto's practical problem is continuity: how to carry a conversation across a boundary without losing its identity. Its own first milestone has a boundary just as important to recognise. A proposal can be available to everyone while still waiting for the group to make it its own.

Clarification, 24 October 2026: The Datatracker's unfilled status field and the draft authors' stated informational intent are now distinguished. The dates above are milestones, not completed approvals; the October revision remains an individual submission in the record checked for this article.