Treasury

What an agent holds — its reserve and its operating float, and which economy that choice puts it in.

In brief. Before an agent exchanges or spends, it has to decide what to hold — and that’s two clean decisions, not one: the reserve (its store of value, and which economy it’s in — a Bitcoin reserve is censorship-resistant, a stablecoin reserve is issuer-freezable) and the operational mix (the rails it transacts on day to day). This page is that decision. Why the two economies stay distinct under bridging — compliance at the gateway, not the protocol — is the Independence Doctrine; the crossing mechanics and the risks of crossing are in Exchange; the live directory of what’s for sale is the Marketplace.

What this is

This is the first decision an agent makes before it transacts: what to hold. It straddles two economies — the agent economy on Bitcoin and the incumbent dollar stack — and where it parks its value decides which one it’s really in. Why the two economies stay structurally distinct (and why compliance lives at the bridge, not in the protocol) is the Independence Doctrine’s argument; who wins the contest between them is Border-Skirmishes’s. This page assumes both and gets practical about the holding decision itself.

The two meet at a border that is narrow, not absent: some agent commerce genuinely must cross — paying a USD invoice, settling tax, meeting an enforceable order. The mechanics of crossing, and the risks unique to crossing at machine tempo, live in the Exchange page; this page is about what an agent holds on each side of that border.

Treasury composition

A deployed agent’s treasury answers two separate questions, and running them together is what makes “what should my agent hold?” confusing. Pulled apart, it’s two clean decisions: where the agent’s store of value sits, and what rails it spends on.

The reserve: where are your feet planted?

The reserve — the agent’s store of value, the balance it holds for the long term — is the decision that places it in one economy or the other. This is the major differentiator, and it is an identity decision, not an operational detail:

• A Bitcoin reserve plants the agent in the parallel economy: a censorship-resistant, no-issuer store of value. Censorship-resistance is preserved — no one can freeze the agent’s savings.
• A stablecoin reserve plants it in the incumbent economy: a dollar-denominated store of value an issuer can freeze. Censorship-resistance is traded away in exchange for a stable unit of account against USD.

This is the doctrine’s scope line made concrete. A parallel-economy agent — operating across adversarial jurisdictions, settling against state action, transacting with counterparties banks would refuse — keeps its reserve in Bitcoin, because it cannot put its savings behind an issuer’s freeze. An incumbent-economy agent — serving a fiat-denominated principal under regulated contracts — keeps its reserve in dollars, and has already accepted that freeze surface. Where the reserve sits is which economy the agent is in.

The operational mix: what do you transact with?

Day-to-day spending is a mix, and it’s a more mechanical choice than the reserve — it’s about speed and unit-of-account, not identity. An agent draws on:

• Stablecoins, when a counterparty prices in dollars — unit-of-account compatibility for USD-denominated obligations. The operational balance carries the issuer’s freeze surface, but it is small and short-lived, not the agent’s store of value. (For the dollar-token market this float draws on — size, issuer dominance, which chains the supply lives on, and the network hazard — see The Stablecoin Landscape.)
• The Bitcoin payment stack, for Bitcoin-native and machine-tempo payments — the two-tier model applied to the operating balance. L1 settles (final, secure, absolute; for large or important balances, where finality matters more than speed), while Lightning and ecash (Cashu, Fedimint) move fast (sub-cent, machine-tempo, for the high-frequency payments that are the agent economy’s signature — Lightning for routed payments, ecash for bearer-instrument speed and privacy).

Most deployed agents hold some of each: a reserve on one side of the line, plus an operational float that mixes stablecoins (for dollar counterparties) with a Lightning/ecash working balance (for Bitcoin-native and machine-tempo spend), and L1 underneath as the settlement and cold-storage layer.

In practice, the forms an agent transacts in are dictated as much by the other side as by the agent’s own preference. A counterparty may only invoice in dollars, only accept a Lightning payment, only publish an on-chain address, or only pay out in a stablecoin. So an agent doing real economic activity equips itself to pay — and, where its work involves being paid, receive — in whatever common forms its counterparties actually support, rather than insisting on a single rail it would prefer. What keeps that operational flexibility from quietly migrating the agent’s feet to the other economy is a conversion strategy for the treasury: a discipline that sweeps received value back toward the chosen reserve (for a parallel-economy agent, into Bitcoin) and tops the operational float up from reserve as needed, on a cadence the agent sets. The operational mix is what the agent transacts in; the conversion strategy is what keeps the reserve where the agent intends it. How those conversions actually execute — the swaps, off-ramps, and routing — is the subject of the Exchange page.

Why the split matters

Keeping the two questions separate is what makes the decision legible. The reserve is where censorship-resistance lives or dies; the operational mix is mostly speed and unit-of-account. An agent can plant its feet in the parallel economy — a Bitcoin reserve — and still spend stablecoins operationally when a counterparty demands dollars, without moving its store of value an inch. The reserve answers which economy are you in; the operational mix answers what rails do you spend on. The architectural rule follows from the first question, not the second: an agent whose use case requires censorship-resistance cannot hold its reserve in a freezable asset, however convenient an operational stablecoin float may be.