wtf is agent-economy?

agent-economy is a working demonstration of an AI agent that gets a real budget and has to spend it wisely. The project gives one AI agent a wallet holding one US dollar in USDC, a type of stable digital currency, and sends it out to shop a small marketplace of paid data tools. The agent has to judge whether each tool is worth the asking price, skip the ones that are overpriced, and even hire a second AI agent to do work it cannot handle itself. Every purchase happens on a real blockchain, using a payment method called x402 that lets a web server simply reply with a payment request instead of a login form. The buying agent, called Scout, follows a strict spending policy and cannot go over its budget or its per purchase limit. In the example run described in the README, Scout rejects a bundle priced at 75 cents because it breaks the 25 cent cap, then buys cheaper, useful data instead. The project is organized into separate pieces. An agent folder holds Scout's buying logic and its own wallet. A marketplace folder runs three sample data tools, one of which is deliberately overpriced to test Scout's judgment. An analyst folder runs a second AI agent, Analyst-7, that sells its own service and gets paid directly into its own wallet, so this is agent to agent commerce, not just agent to shop. A dashboard folder shows a live view of the budget, the decisions made, any rejections, and receipts recorded on chain. To try it, you install dependencies in each folder, create wallets, fund them with free test USDC, then run each piece in its own terminal window. A full demo run spends about 40 cents and finishes in under two minutes. The project uses Claude as the judgment layer that decides what to buy, skip, or hire. It currently has no stars and is written mainly in JavaScript.