AI x402 Paying Agent

This open-source app is an AI agent that carries its own crypto wallet and answers a plain-English question by paying for the data it needs through the x402 standard, aiming to avoid API keys, subscriptions, and signup.

Why it caught our eye

Rather than relying on pre-provisioned API keys, the agent reacts to a 402 Payment Required response by having the x402 SDK automatically sign a USDC payment, with a per-call price ceiling intended to cap overspending.

What it does

The agent takes a plain-English question. Claude decides which paid API should answer it and calls a fetch tool. When an API replies with 402 Payment Required and a price quote, the x402 SDK automatically signs a USDC payment and the agent pays. This is the documented workflow; in this introduction we describe it rather than report having run it ourselves.

A concrete possibility

If you want to watch the payment flow on its own, a --direct mode pays and fetches any x402 URL with just a wallet and no LLM, isolating the paid-fetch step from the question-answering layer.

Intended users and access

It is aimed at developers who want to see an agent handle per-call payments and who are willing to set up wallets and an Anthropic API key. Running it involves cloning the repo, installing requirements, configuring buyer and seller wallets plus an Anthropic API key, and starting the bundled FastAPI seller via uvicorn. The documented demo is testnet-only: payments settle in testnet USDC on Base Sepolia through the public x402.org facilitator, and the buyer wallet is funded from a free faucet rather than real money.

Known limitations

The documented demo uses testnet USDC with a faucet-funded wallet, so it is not presented as a real-money deployment. Spending is constrained by a per-call price ceiling (MAX_PRICE_USDC, default $0.05), which limits each call but is a per-call cap rather than stated to be a total budget. The materials do not provide pricing, release dates, or evidence of a production or mainnet rollout.

Sources & freshness

  1. raw.githubusercontent.com Checked 2026-10-23

AI-assisted introduction based on public sources. We have not conducted a hands-on review. Check the official site for current access and pricing.