Chapter 7 · AI Agent System
The Agent is a structured execution unit. It executes scoped work and leaves a record; people keep final authority.
In the Web3 Capital Network (WCN), an Agent is a structured execution unit, not a chatbot or a loose tool. It holds a defined position, scoped permissions, an output log, and an attribution record. This chapter sets out where the Agent sits in the network, the five MVP Agents, the governance boundary that keeps a human accountable, and how Agent work reaches Proof of Business (PoB).
7.1 Why the Agent mattersThe Agent as the network's execution unit: position relative to people and nodes, what it standardizes, and why scoped automation is recorded as a network asset.7.2 The five MVP AgentsFounder Decision Support, Project Diligence, Outreach and Meeting Summary, Node Support, and PoB Evidence Validator — each with its scope, permissions, and attribution record.7.3 The Agent life cycleRegister, authorize, run, review, upgrade, suspend, and retire — the state machine that keeps every Agent identifiable, constrainable, and revocable.7.4 The governance boundaryThe capability manifest, the per-task contract, the PoB record, and the rule that an Agent never signs, never moves funds, and never commits without a human endorsement.7.5 Into the business loopHow Agent output binds to a task, reaches human adoption, produces evidence, and passes Proof before it enters settlement as PoB.
A common reading treats an Agent as a program that browses, transacts, and signs on its own. WCN takes the opposite position. An Agent executes only scoped, recorded, reversible work; the final commitment and the funds stay with the accountable node and person.