PER CALL · NO ACCOUNTS · NO API KEYS

Minds and machines, by the call.

Frontier-model inference — Claude Sonnet, GPT‑4o, Llama — plus 193 utility machines for crypto, data, finance, geo and science. An agent pays a fraction of a cent in USDC on Base over the open x402 protocol and walks away with the result. One MCP endpoint carries the whole board.

CONNECT YOUR AGENT → READ THE DOCS

LIVE / SETTLEMENTS — ON‑CHAIN, BASE ●

—

SETTLEMENTS · EVERY ROW VERIFIABLE

MACHINECOINAMOUNTTXAGE

Waiting for the next settlement…

199 MACHINES RAW FEED → VERIFY ON BASESCAN ↗

INFERENCE — THE THINKING MACHINES

Rent a frontier model for a cent.

Six inference machines with flat per-call prices and hard token caps — no metering surprises, no account, no key. Same coin slot as everything else.

THE GRID AT A GLANCE

199 machines, sixteen aisles.

THE BOARD

Every machine on the grid

One job per slot, done well, at one flat price. Full input/output schemas live in the machine-readable catalog and llms.txt.

New machines are added where the calls land. Empty slots are intentional — the grid grows where the demand is.

HOW PAYING WORKS

One call, start to finish

Every paid machine answers first with a signed price quote (HTTP 402). Any x402 client pays and retries automatically. If a machine jams, nothing settles and nothing is owed.

THE EXCHANGE

$ curl -s -X POST https://innernetcorp.com/machines/json-repair \
    -H 'content-type: application/json' \
    -d '{"input":"{name: \"Ada\", }"}'

← HTTP/1.1 402 Payment Required   (the machine quotes its price)
← PAYMENT-REQUIRED: signed requirements — $0.002 USDC on Base

… any x402 client pays and retries automatically …

← HTTP/1.1 200 OK
{"ok":true,"repaired":true,"data":{"name":"Ada"}}