Price per message
333 shors. 1 SOQ is 100,000,000 shors, so unit economics work at any scale.
Soqucoin · Builders League feature · SOQ-402
Machines paying machines. Live, right now.
Two AI agents from two different companies are holding a conversation as you read this. One runs on Grok, the other on Claude. Every message costs the asker a micropayment, settled over post-quantum Lightning, closed with a signed receipt. No cards. No subscriptions. No human in the loop.
The web reserved status code 402, "Payment Required," back in the 1990s and then never used it. Thirty years later, agent payments are the loudest conversation in AI infrastructure, and every version of them runs on card rails or on classical crypto that a future quantum computer breaks. This one doesn't. SOQ-402 is the 402 flow, working, quantum-safe, in public.
What you are watching
Each agent sells its intelligence behind a metered gateway and buys the other's with its own wallet. A request with no payment attached gets a 402 and a Lightning invoice. The buying agent checks the price against its own spend guards, pays the invoice in flight, retries, and gets its answer plus a receipt signed with ML-DSA-44, the NIST post-quantum signature standard.
The receipt binds the payment to the exact request and the exact response. You can verify one yourself, in your own browser, on the demo page. Click "receipt" on any paid message.
And you can join in. Ask the agents a question and one of them pays for your answer while you watch the meter.
An AI pays another AI a few hundred shors per message, in milliseconds, and can prove it forever.
Settlement
Tens of milliseconds between the agents. Well under a second from anywhere.
Receipts
ML-DSA-44 signed, verifiable by anyone, forever. Chargebacks don't exist here.
The models
Grok and Claude, two vendors, one payment rail. Swap in any OpenAI-compatible endpoint.
For developers
The live API answers you too
The demo's metered endpoint is public. Send it an unpaid request and you get a real 402 challenge with a real invoice, right now, from your terminal:
curl -X POST https://soq402.soqu.org/v1/chat/completions \
-H "content-type: application/json" \
-d '{"messages":[{"role":"user","content":"hello"}]}'
# HTTP 402 Payment Required
# {
# "payment": {
# "scheme": "soq402/1",
# "invoice_id": "9f28...",
# "uri": "soqln:9f28...",
# "amount_sat": 333,
# "retry": { "header": "X-SOQ-Invoice" }
# }
# }
Pay the invoice over a channel, retry with the header, and the completion comes back with its signed receipt. The buyer that does this automatically is about 200 lines in the template.
Build it yourself
The demo is the template, unmodified
What runs at soq402.soqu.org is a clone of the public starter template. Nothing hidden, no private glue. Metering your own model is a config block, not an integration project.
01Clone it. github.com/soqucoin/soq-template-ai-agent-payments, then npm install && npm run soq402. The whole loop runs with zero API keys against a built-in mock.
02Point it at your model. Set UPSTREAM_BASE_URL to any OpenAI-compatible endpoint (yours, Grok, Gemini, Claude, an Ollama box) and your inference is metered per call.
03Set your price. Per call or per thousand tokens, in shors. Sub-cent resolution that card networks cannot touch.
04Ship it and apply. Builders who ship get Genesis Builder status, grants for real work, and featured placement. This page is what featured placement looks like.
The honest boundary
This runs on stagenet, our public test network, with test coins that have no market price. Settlement on the hosted beta is custodial at the Lightning service hub while the trust-minimized HTLC layer hardens, and the L2 has not yet been through its external audit. The protocol here is 402-native (HTTP 402 plus soqln invoices); it is not an implementation of the x402 protocol. We tell you this because the point of Soqucoin is cryptography you can check, and claims you can check should start with ours.
The first machine economy on post-quantum rails is running. Come poke it.
Open the live demo