Work coordination layer
The system that records what work exists, who or what is doing it, and what state it is in — above the tools that execute the work.
The work coordination layer answers three questions for an organisation: what needs doing, who or what owns it, and where it stands. Editors, terminals, browsers and CLIs sit below it as execution surfaces — where a task is carried out, not where it is tracked. Conventional project trackers sit in this layer but assume a human executor, so an agent's run is at best pasted back in as a comment.
The layer matters more as agents get more autonomous, because autonomy without coordination produces duplicated and unowned work: two people prompting two agents for the same thing, with nothing recording that either run happened. LangChain's State of Agent Engineering reports the same shift from a different angle — as teams put agents into production, the binding constraints become observability and control rather than model capability.
This is a descriptive term coined for that layer, not established industry vocabulary.