
Agent Economy is a budget-aware AI agent pipeline that demonstrates machine-to-machine commerce using USDC nanopayments on Arc testnet. When a user submits a task like "Analyze crypto sentiment with budget $0.01", the system autonomously plans a multi-agent pipeline, selects the most cost-efficient APIs within budget, executes each step, and settles every API call as a real USDC transaction on Arc — all in under 10 seconds. HOW IT WORKS: The Planner parses natural language input and determines which agents to activate. The Decision Engine scores available APIs using quality/price ratio and rejects any that exceed the remaining budget — logging its reasoning transparently. Three specialized agents (DataAgent, AnalysisAgent, ReportAgent) each call their assigned APIs and trigger micropayments through the Payment Engine. Every payment is a real ERC-20 USDC transfer on Arc testnet, confirmed onchain before the next step proceeds. KEY METRICS FROM LIVE DEMO: - 52 onchain transactions confirmed on Arc testnet - Average cost per transaction: $0.000515 USDC - All transactions under $0.01 (nanopayment threshold) - Total demo spend: $0.0268 USDC - Gas overhead vs ETH mainnet: 1940x cheaper WHY THIS ONLY WORKS ON ARC: On Ethereum mainnet, gas costs $0.50–$5.00 per transaction. Our average payment is $0.000515 — meaning gas alone would be 1000x the payment value, making the entire model economically impossible. Arc's USDC-native gas with sub-cent fees is the only reason per-API-call settlement is viable at scale. THE WEB UI: A live web dashboard streams every agent decision, API selection, payment event, and transaction hash in real time. Users can submit tasks, watch the pipeline execute, and verify every transaction directly on the Arc testnet explorer (testnet.arcscan.app). This project demonstrates a concrete, working implementation of the agentic economy — where AI agents earn, spend, and settle value autonomously without human intervention.
26 Apr 2026