Payment and permission checks for agent-initiated work

IETF Datatracker · Individual Internet-Draft ·

Sets out a server-side model for work that AI agents commission and receive later. Defines the task lifecycle, places checks on delegated authority and payment or entitlement before fulfilment, and links repeated requests to one task.

Read the Internet-Draft on the IETF Datatracker (37 pages)

The document

Title
A Server-Side Model for Gating Agent-Initiated Human-Sourced Asynchronous HTTP Tasks on Payment or Entitlement
Identifier
draft-king-yew-choo-agentic-payments-00
Intended status
Informational
Posted
Expires
Length
37 pages
Adoption
An individual submission. Not adopted by any IETF working group.

What the model does

It describes paid, human-sourced asynchronous HTTP tasks initiated by software agents acting for a principal. Clients may retry after uncertain outcomes and act under authority delegated in advance, while fulfilment incurs cost and may not be freely reversible.

The model places a payment or entitlement gate before fulfilment, maps that gate to adjacent protocols, and defines a task state machine.

What it yields

  1. Gate-before-fulfilment safety. No fulfilment ahead of the check.
  2. At most one task per retained idempotency key. A retry under the same key does not create a second task.
  3. At most one gate acceptance per payment requirement. A single requirement cannot be satisfied twice.
  4. An operator-checkable record. Linking a delivered artefact to operator-controlled records.

All four hold under stated well-formedness constraints and operating assumptions. The draft defines no protocol, no wire format and no conformance requirements.

Sources