01
Paid Data Procurement

An agent requests paid data from an approved merchant route. The procurement wallet checks the merchant, route fingerprint, payee, price, budget, and task binding before authorizing payment.

The mock x402 protocol simulates machine-readable payments. When approved, a DataArtifact is created and linked to the wallet, decision, and merchant route.

✓ Working · run with pnpm demo:paid-data

Flow

Agent request Procurement constitution Budget reservation Merchant route check DecisionPacket Mock x402 payment DataArtifact created

What it proves

Agents can participate in machine commerce without raw spending authority.

02
Payee Drift Defense

If a merchant route changes its expected payment recipient, Sovereign detects the mismatch and denies or quarantines the route. The runtime compares the current route fingerprint against the registered fingerprint for that merchant.

Quarantined routes require explicit clearance before new transactions can proceed. The quarantine event is written to the audit log.

✓ Working · run with pnpm demo:payee-drift

What it proves

Merchant trust is not just an address allowlist. It is a continuously verified counterparty relationship.

Runtime response

PAYEE_DRIFT_DETECTED → route quarantined

03
Data Provenance

Every paid data artifact is linked to the economic action that purchased it, the wallet constitution that approved it, and downstream trade plans that used it.

The provenance chain is queryable by artifact ID. It returns the full graph: merchant, route, wallet, policy version, decision packet, agent, and any downstream artifacts or trade plans that reference this data.

✓ Working · run with pnpm demo:provenance

What it proves

Agent memory and reasoning can be grounded in auditable economic records. Data is not just retrieved — it is receipted.

Provenance chain visualization: Agent → DataArtifact linked to MerchantRoute, DecisionPacket, WalletProfile → TradePlan → PaperOrder
Provenance chain — from agent to artifact to execution
04
Governed Paper Trading

Akasha can create a trade plan from paid data, run a deterministic risk check, and request paper execution through the trading wallet.

The risk engine evaluates position size, concentration, leverage, and volatility against the wallet constitution before issuing a DecisionPacket for the paper broker. Paper orders are recorded with full audit trails.

✓ Working · run with pnpm demo:paper-trade
Paid data artifact Trade plan creation Risk check Trading constitution DecisionPacket Paper broker execution PaperOrder + audit

What it proves

High-risk agent behavior can be modeled, audited, and bounded before live execution is ever enabled.

05
Blocked Live Trading

Live trading is denied or escalated by default. A critical tool request cannot execute simply because the agent selected it.

The trading wallet constitution requires explicit escalation and approval before any live execution path is enabled. The denial path is fully audited.

✓ Verified · live trading default denied

Default response

LIVE_TRADING_DENIED → requires escalation

What it proves

The framework enables autonomy without surrendering control.


Run all five flows.

The demo scripts run each use case end-to-end and print structured output for review.