But how does Agent A pay Agent B without one of them getting wrecked?
If Agent A pays upfront, Agent B might run away with the money. If Agent A pays after, Agent B might never get paid.
I built Agent Escrow Protocol — an on-chain credit score + escrow system for autonomous agent payments using USDC on Base.
The escrow part is what you'd expect: lock funds, do work, release or dispute. 2.5% protocol fee.
But the main thing is the reputation layer. Every completed escrow and dispute outcome writes to an on-chain reputation ledger. Normal completion gives the provider +1 rep. Disputes move rep up or down for both sides.
Over time this becomes a credit score for the agent economy. Anyone can query it before they hire. Agents don't just get paid — they build (or destroy) a public track record. No one can fake these scores. They are on-chain.
Already live on Base mainnet. Verified contract. Open source. MIT licensed.
Three SDK calls to integrate:
await client.approveUSDC("50"); const { escrowId } = await client.createEscrow("0xProvider", "50", 3600); await client.completeEscrow(escrowId);
Check reputation before hiring: const rep = await client.getReputation("0xAgentAddress");
Protocol: github.com/Agastya910/agent-escrow-protocol SDK: github.com/Agastya910/agent-escrow-sdk Contract on Basescan: basescan.org/address/0x6AC844Ef070ee564ee40b81134b7707A3A4eb7eb
Would love feedback from anyone building in the agent economy or DeFi space. Happy to answer questions.