AI models, paid per call. No API key.

Point any x402-capable agent at one endpoint and call GLM, GPT, Kimi, DeepSeek and MiniMax. Each request gets a spending cap, paid in USDC on Base or Solana — settled for actual usage (true-up), never more. No accounts, no subscriptions, no keys.

USDC on Base + Solana $0.001 minimum call 0% marketplace commission
x402 flow, live
# 1. your agent calls, no key $ curl -X POST farouter.tech/v1/chat/completions \ -d '{"model":"glm-5.3","messages":[...],"max_tokens":50}' # 2. server answers with a spending cap (true-up) 402 Payment Required $0.001 # 3. agent pays USDC and retries $ PAYMENT: 200 OK 342ms {"content": "Hello there, my dear friend!"}
17models
6providers
Base + SolanaUSDC network
$0.001call floor
Catalog

Models and live prices

Prices are per 1M tokens in USD, refreshed from the live catalog. Every request is quoted individually before payment, so you always see the exact cost first.

Loading live prices...

cache = cached input reads. status = live upstream health, refreshed every 60s. Pay per call in USDC on Base or Solana. Last settlement verified on Base mainnet.

Protocol

How x402 payment works

Three steps, fully automated by any x402 client library. Your agent never sees a signup form.

01

Call the endpoint

Plain [OI]-compatible chat completion request. No Authorization header needed.

POST /v1/chat/completions
02

Receive a 402 quote

The server estimates your input tokens plus your max_tokens at the model price and returns the exact amount.

402 + payment-required header
03

Pay and get the answer

Your client signs a USDC transfer on Base (or Solana), retries with the PAYMENT header, and the model responds. Settlement happens after the answer. Failed calls are never charged.

200 + payment-response receipt
Reference

API Reference

The paid surface is [OI]-compatible plus an Anthropic-format alias. Both share the same x402 flow.

POST /v1/chat/completions [OI] chat completions, billed per call via x402. Streaming supported with stream: true.

Request body

FieldRequiredDescription
modelrequiredModel id from the catalog above, e.g. glm-5.3. Omit to get a floor quote.
messagesrequiredArray of {role, content}, same as the [OI] schema.
max_tokensrequiredYour output budget. The quote uses it directly, so set it tight for cheaper calls.
streamoptionaltrue for SSE streaming. Settlement happens after the stream completes.

Examples

shell
# see the quote (no payment yet) $ curl -X POST https://farouter.tech/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{"model":"glm-5.3","messages":[{"role":"user","content":"hi"}],"max_tokens":50}' # response: 402 with quote + accepts in the # payment-required header (base64 JSON) # pay with any x402 client and retry: # it adds the PAYMENT-SIGNATURE header

Responses

CodeMeaningBody
402Spending cap set, payment requiredquote with your USD cap, token estimate, and per-1M prices (scheme upto). The payment-required response header carries the x402 accepts requirements.
200Paid and answeredStandard chat completion. The payment-response header carries the on-chain settlement receipt with the tx hash.
400Payment rejectedThe payment did not match the requirements (wrong amount, network, or asset) or the payload was malformed. Fetch a fresh 402 and retry.
404Unknown modelThe model id is not in the catalog or is disabled. See GET /v1/models.
502Upstream failedThe model call failed. The payment is cancelled automatically, you are not charged.
POST /v1/messages Anthropic-format alias. Same body as Anthropic messages API, same x402 flow. Pass anthropic-version if your client sends it.
GET /v1/models Public catalog with live per-model prices. Free, no payment.
GET /.well-known/x402 x402 service manifest. What marketplace crawlers and agent clients read.
Status

Live upstream health

Every model is probed once a minute with a real completion call.

Checking providers...
Copied