---
title: "OpenAI-compatible chat completions across every frontier..."
method: POST
path: "/api/ai/chat"
tags: ["ai"]
---

# OpenAI-compatible chat completions across every frontier...

`POST /api/ai/chat`

## Request

- Base URL: `https://2s.io`
- URL: `https://2s.io/api/ai/chat`
- Auth: API key in header `PAYMENT-SIGNATURE`

OpenAI-compatible chat completions across every frontier model on one endpoint — GPT-5, Claude, Gemini, Grok, DeepSeek, Llama, Mistral and ~290 more. POST { model, messages, max_tokens?, temperature?, top_p?, stop? } and get back a standard chat.completion (choices[].message + usage). Pay per call in USDC via x402 — no accounts, no provider keys, no subscriptions. Price scales with the model and max_tokens and is quoted in the 402. List selectable models at GET /api/ai/models. Billing: payment settles before this endpoint runs (paid upstream), so an error after settlement IS charged — the error response then carries the payment receipt (PAYMENT-RESPONSE) and `X-2s-Charged: true`. upto-scheme payments settle only after success.

## Request body

- object
  - `model` string, required: Gateway model id, e.g. openai/gpt-5, anthropic/claude-opus-4.8, google/gemini-3-pro-preview. Full list: GET /api/ai/models.
  - `messages` object[], required: OpenAI-style message list.
    - `role` 'system' | 'user' | 'assistant', required: Role.
    - `content` string, required: Content.
  - `max_tokens` integer: Max output tokens (default 512).
  - `temperature` number: Temperature.
  - `top_p` number: Top p.
  - `stop` union: Stop.
    - string
    - string[]

## Response `200`

OK

## Other responses

- `400`: Invalid request. Parameter validation (BAD_REQUEST / INVALID_PARAM with error.details.issues, MALFORMED_JSON) runs before payment and is never charged; a request-specific 400 raised by the endpoint itself runs after settlement on this paid-upstream operation and IS charged (`X-2s-Charged: true` + PAYMENT-RESPONSE receipt).
- `402`: Payment required. Body is the x402 PaymentRequirements envelope with a multi-network accepts array; the per-call price is in accepts[].amount (and on the operation as x-2s-price). Sign for whichever rail you hold USDC on (EIP-3009 for Base, partial SPL transfer for Solana) and retry with the PAYMENT-SIGNATURE header (X-PAYMENT also accepted for v1 clients). Also PAYMENT_SETTLE_FAILED (error envelope) when a verified payment failed to settle: error.details.chargeStatus is "not_charged" (retry with a fresh payment) or "unknown" (the authorization may still settle before its deadline — check it before retrying).
- `409`: Conflict — PAYER_BUSY: another upto call from this payer is still in flight. Not charged; retry after Retry-After, run upto calls from one wallet sequentially, or pay with the exact scheme.
- `429`: Too many requests — RATE_LIMITED: our per-IP / per-payer limiter. Not charged; retry after Retry-After.
- `5XX`: Server-side failure. 500 INTERNAL: unexpected error; quote error.errorId (also the X-2s-Error-Id header). 502 UPSTREAM_UNAVAILABLE (or another UPSTREAM_* code): the data provider failed — composite endpoints refuse unbilled when every source fails. 503 SERVICE_PAUSED: endpoint paused by the operator kill switch, retry after Retry-After; 503 SPEND_CAP: daily upstream budget exhausted. 504 UPSTREAM_TIMEOUT. On a paid-upstream operation (see its description) this can follow settlement: then the call WAS charged and the response carries `X-2s-Charged: true` plus the PAYMENT-RESPONSE receipt. SERVICE_PAUSED and SPEND_CAP are never charged.

---

[API](https://skmtc.dev/2s-io/apis/2s-the-most-everything-api.md) · [All operations](https://skmtc.dev/2s-io/apis/2s-the-most-everything-api/llms.txt) · [OpenAPI document](https://skmtc.dev/2s-io/apis/2s-the-most-everything-api/revisions/da4c5caf686e?raw)