x402 Mesh Inference

Pay-per-prompt AI inference. An agent calls an ordinary OpenAI-compatible endpoint with no account and no API key, pays $0.002000 USDC inline over HTTP 402, and the prompt is routed to an independent GPU operator who is paid on chain.

network testnet USDC ASA 10458941 scheme exact x402 v2 status checking…

Live mesh

Nodes online
—
Models served
—
Requests settled
—
Paid to operators
—

What one request costs

USDC has 6 decimals, so every figure below is an exact integer of atomic units. The gateway never does floating-point arithmetic on money, and asserts inbound − payout = margin before any funds move.

LegDisplayAtomic
Client → gateway$0.0020002000
Gateway → node operator$0.0017001700
Gateway margin (1500 bps)$0.000300300

How a payment happens

1
Agent POSTs an OpenAI-shaped body with no payment header.
2
Gateway answers 402; the machine-readable challenge rides in the payment-required header.
3
Agent retries with PAYMENT-SIGNATURE — an Algorand atomic group. The facilitator sponsors the fee, so it is gasless for the payer.
4
Facilitator verifies, then the healthiest node serving that model gets the prompt.
5
Completion streams back over SSE while settlement lands on chain.
6
On confirmed settlement only, the operator is paid their share.

Try the handshake

This sends a real unpaid request to the paid route and shows you the actual challenge that comes back — no wallet required.

integrate in 60s → or open the chat client →

For agents

/.well-known/x402 /llms.txt /v1/pricing /v1/nodes /v1/settlements /readyz