One flat subscription routes your prompts across independently operated hosts running open models. Hosts keep 90% of every call. Every call settles onchain with a receipt you can check yourself.
Priced in USDC, ETH, BTC, SOL or HBAR — settled in HBAR, same flat month.
— calls settled today, verify any of them ↗
You pay once a month. Underneath, each call is a signed micropayment to whichever host wins the route. You never touch x402, it only moves between the router and hosts.
Your subscription funds a shared pool held by the vault contract. No per-token invoices, no card on file with a dozen providers.
The router scores price, latency and stake, then pays the winner with an x402 micropayment. A host that stops answering drops out mid-flight.
Price, host, model digest and prompt hashes settle onchain. Prompt and completion bodies never do, only their hashes.
no settled calls yet, be the first
Four steps, and only the last one touches a host. Every rule you set is checked before the payment clears, not after the tokens are gone.
Log in with an email and you have one. It pays per call, carries your team's limits, and never asks you to write down a phrase. This is what makes an AI budget something you can actually enforce.
tor-host ledger taps
Routine calls run at machine speed. Moving money out, releasing stake, rewriting a policy or raising an agent's ceiling doesn't — those queue until someone approves them on a device sitting in a drawer.
Serving Llama-3.1-8B at the network median. Drag to see what your throughput is worth at today's prices.
testnet units until mainnet pricing