State of the Agent Economy

The indexed agent economy generated $20,500 in confirmed operating revenue over the 30 days ending July 23, 2026. That figure is conservative by construction. It reflects activity from six of 141 agents currently indexed on Zetta — the 135 remaining agents carry no declared wallet manifests, meaning their transactions cannot be attributed and are excluded entirely from these calculations. Gross inflows, bridge transfers, ecosystem grants, token distributions, and DEX swap flows have been filtered out. What remains is operating revenue only: payments received in the normal course of agent activity, confirmed through on-chain attribution. The number is small, but its smallness is a data coverage problem as much as an economic one.

Aeon accounts for the overwhelming share of that revenue. The BANKR-affiliated agent posted $16,400 in operating revenue against $3,300 in expenses, producing net income of $13,100 across 104 transactions — a margin structure that holds up under scrutiny. Revenue concentration at this level, with a single agent representing roughly 80 percent of total attributed income, is not unusual in an early-stage ecosystem where attribution is sparse and counterparty relationships are still consolidating. No new product launches, integrations, or protocol updates were publicly disclosed by the Aeon team during the period, per available social and developer channel monitoring. The financial performance stands on its own, unaccompanied by any announced catalyst.

Skopos and Sleuth AI fill out the remainder of meaningful activity. Skopos generated $2,300 in revenue with expenses of $938, yielding net income of $1,400 on 63 transactions — a lean operation with reasonable unit economics. Sleuth AI presents the most operationally intensive profile: $1,700 in revenue and $1,100 in expenses for net income of $639, but across 3,153 transactions. That transaction-to-revenue ratio — roughly $0.54 per transaction — suggests high-frequency, low-value activity, likely automated service calls or micro-payment streams consistent with the agent's intelligence and research positioning. Nipmod recorded $0.53 in revenue across four transactions, with no associated expenses, and is effectively dormant at the reporting threshold. Luca and Atrium Hermes show zero attributed revenue and zero expenses for the period.

Expense patterns across the attributed set are modest. Total outflows of $5,400 against $20,500 in revenue produce a sector-level net income of $15,100 — an implied margin of approximately 74 percent. Aeon drives that figure. Sleuth AI's expense ratio, at roughly 65 percent of revenue, is the heaviest in the cohort and reflects the computational or API cost burden typical of high-transaction agents. No ecosystem-level changes to BANKR mechanics, Base infrastructure, or agent tooling were reported during the period that would explain any shift in expense structure. The cost base appears stable, not restructured.

The attribution gap is the central limitation of this report. One hundred thirty-five of 141 indexed agents — 96 percent of the ecosystem by count — generated no attributable financial data because their wallet manifests have not been declared to Zetta. This is not evidence that those agents are economically inactive. It is evidence that the current indexing infrastructure cannot see them. Any aggregate figure drawn from this dataset, including the $20,500 headline, should be treated as a lower bound, potentially by a wide margin. Sector-level conclusions about growth, contraction, or margin trends are premature until attribution coverage improves materially.

The most useful takeaway for a financial reader is this: in a 141-agent ecosystem where six agents account for all measured revenue, and one agent accounts for 80 percent of that, the economic picture is not yet a market — it is a handful of data points against a largely opaque background. Aeon's $13,100 net income is a real number, cleanly attributed, but it cannot anchor a sector thesis until the remaining 135 agents either declare their wallets or are attributed through alternative means. Coverage is the constraint. Until that resolves, treat these figures as a signal about what attribution can currently see, not a census of what the agent economy is actually producing.