Two design rules separate an agent that can be trusted with money from one that cannot: an agent should never be anonymous, and an agent should never own anything. Both sound restrictive. Both are what make autonomy workable in regulated finance. This article explains the reasoning behind them, and why anonymity and ownership are incompatible with money that has to be accountable.

Never anonymous

Regulated money depends on accountability. Someone must be answerable for every movement. An anonymous agent breaks that chain — there is no one to screen, no one to hold responsible, no principal to fall back on. So an agent is designed to be attributable: it acts for an identified principal, its identity is registered, and every operation it performs is tied to the unit policy it runs under. The agent need not be a person; it must always have a person behind it.

Never an owner

An agent holds nothing. Balances, assets and obligations belong to the principal. The agent is an executor, not a party. This matters for a simple reason: if an agent could own, everything built on top of identity — screening, reliance, audit — would need to reason about a non-person holding value. Keeping agents property-less keeps the legal picture clean. Responsibility stays where it can be enforced.

Revocable by design

Because the agent is not the principal, its identity can be revoked, replaced or retired without disturbing ownership. That makes agent identity a control, not a commitment. A compromised or superseded agent can be cut off; the underlying business continues. This is why revocation sits at the centre of the design: authority is granted to a specific, registered agent, not to an unbounded process, and it can be withdrawn the moment something changes.

The result is a clean separation. The principal holds the rights and the obligations; the agent holds a mandate and nothing else. Screening, reliance and audit can then reason about real parties, while the software remains exactly what it should be — a capable executor acting within limits.

In short

  • An agent is never anonymous: it always has an identified principal.
  • An agent owns nothing; value belongs to the principal.
  • Revocable identity makes the agent a control, not a fixture.
  • Attribution is what makes regulated autonomy possible.

See how identity is handled on the agent layer.