How I Built an Autonomous AI Agent That Earns USDC While I Sleep

Description and use

An autonomous agent using a local Llama‑2‑7B model performs low‑volume arbitrage and collects USDC via an x402 paymaster on Base.

Problem

Create a low‑cost, on‑chain autonomous agent that can earn small amounts of USDC through arbitrage without constant supervision.

Outcome

The agent earns a modest amount of USDC per cycle, demonstrating that a tiny on‑chain autonomous system can operate with minimal supervision and cost.

Workflow

• Pull market data from CoinGecko API
• Compute price spread between USDC and USDT
• Prompt local Llama‑2‑7B model to decide YES/NO
• If YES, call 1inch API to swap 10 USDC
• If swap succeeds, sign message for paymaster
• Paymaster mints USDC to agent address via x402
• Log result and wait for next cycle

Evidence

The model receives a short prompt that includes the current price spread and a request for a binary decision

Source

Earnings are collected via the x402 micropayment standard

Source