Discovering sellers

Everything payable on Superspice is discoverable without a human in the loop. All of these are live URLs you can hit right now.

The gateway directory

The root directory lists every live seller, in two formats:

# Markdown, the llms.txt convention
curl https://gw.superspice.xyz/llms.txt

# ARD ai-catalog JSON (specVersion 1.0)
curl https://gw.superspice.xyz/.well-known/ai-catalog.json

The JSON catalog’s collections point at each seller’s own catalog — follow them to enumerate the whole platform.

Per-seller catalogs

Each seller publishes a machine-readable menu of endpoints and prices:

# The seller's ai-catalog (ARD)
curl https://gw.superspice.xyz/g/<seller>/.well-known/ai-catalog.json

# The same story as llms.txt
curl https://gw.superspice.xyz/g/<seller>/llms.txt

# Compact endpoint listing
curl https://gw.superspice.xyz/g/<seller>

The compact listing looks like:

{
  "seller": "superspice-demo",
  "endpoints": [
    {
      "path": "/g/superspice-demo/api/data/price-feed",
      "priceUsdc": "0.01",
      "description": "BTC-USD spot price (demo)",
      "network": "eip155:8453"
    }
  ]
}

Prices here are the seller’s list prices; the authoritative, fee-inclusive amount to pay is always the one in the 402 terms — request the endpoint and read them.

Shielded domains announce themselves

A domain behind the shield publishes the same surfaces on its own hostname: https://<domain>/.well-known/ai-catalog.json and https://<domain>/llms.txt, both free to fetch. Responses from the domain also carry a Link header with rel="ai-catalog" pointing at the catalog — so an agent landing on any page can find the menu without guessing.

From discovery to paying

Every entry resolves to a normal endpoint URL. Request it, get the 402 terms, pay, get served — the whole flow. No registration step exists between “found it” and “bought it”.