I built a machine-payable crypto intelligence API.
No API keys. No subscriptions. No billing dashboard.
AI agents pay per request in USDC using HTTP 402.
How It Works
- Agent calls endpoint
- Server responds with HTTP 402 Payment Required
- x402 protocol handles payment in USDC on Base
- Request retries automatically
- Data returns
The API becomes its own billing system.
What's Inside
- 819 agentic workflow prompts — ready-to-run automations
- Kronos intelligence — hourly crypto signals, risk state, directional decisions, post-trade audits
- $0.05 USDC per call — fractions of a cent to a few cents
Every decision gets a decision_id. Every decision can later be audited against real market data.
Setup
npx coinopai-mcp
Or add to your mcp_config.json:
{
"mcpServers": {
"coinopai": {
"command": "npx",
"args": ["coinopai-mcp"]
}
}
}
Fund a burner wallet with a few dollars of USDC on Base and your agent can start making paid requests immediately.
The Interesting Part Isn't "Signals"
It's verifiable machine-to-machine commerce.
A full decision loop:
preflight → decision → audit
The audit endpoint exists because predictions fail. The system records that too.
Most AI systems quietly hide failures. I wanted the opposite:
- Persistent audit trails
- Explicit verification
- Observable outcomes
Built on x402
The x402 open protocol is gaining serious traction:
- 22 founding members including Visa, Google, AWS, Stripe, Coinbase, Cloudflare
- $600M annualized volume
- Circle Agent Marketplace launched May 11, 2026
- AWS Bedrock AgentCore Payments announced same week
USDC is the default currency of the machine economy. Base mainnet is where the action is.
Try It
- npm: npmjs.com/package/coinopai-mcp
- Dashboard: coinopai.com
- Source: built on x402 + Base + USDC
This is experimental infrastructure. Not financial advice. Predictions fail — that's why the audit endpoint exists.
Top comments (0)