2 · Your first request

The wire format is OpenAI's. Any client that can take a base URL works unchanged — the official openai SDKs, LangChain, LlamaIndex, Codex CLI, Cline, Aider.

curl

curl https://api.tokentoken.ai/v1/chat/completions \
  -H "Authorization: Bearer $TOKENTOKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tt/kimi-k3.1-pro",
    "messages": [{"role": "user", "content": "Say hello in five words."}],
    "max_tokens": 64
  }'

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://api.tokentoken.ai/v1",
    api_key=os.environ["TOKENTOKEN_API_KEY"],
)

resp = client.chat.completions.create(
    model="tt/kimi-k3.1-pro",
    messages=[{"role": "user", "content": "Say hello in five words."}],
    max_tokens=64,
)
print(resp.choices[0].message.content)
print(resp.id)            # keep this: it is the id on your invoice line

Node

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.tokentoken.ai/v1",
  apiKey: process.env.TOKENTOKEN_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "tt/kimi-k3.1-pro",
  messages: [{ role: "user", content: "Say hello in five words." }],
  max_tokens: 64,
});

Streaming: pass stream: true. You do not need stream_options for usage — we obtain usage from the upstream ourselves, so billing never depends on the client asking for it.

4 · How the charge is computed

Prices are published per million tokens. The charge for one call is an exact integer computation — we do not use floating point anywhere in the billing path, and we do not use the gateway's own cost estimate, because it ignores cache discounts.

charge = round_half_up(
    ( prompt_uncached × input_price
    + cached_tokens   × cache_read_price
    + completion      × output_price ) × rate
)

Reading your usage back

GET /v1/pricing                  → { rate, rate_version, currency, models: [...] }
GET /v1/usage?from=&to=          → one row per call: request_id, model, tokens, charge, rate_version
GET /v1/credits                  → { balance, currency, as_of }

Every row carries the request_id that the API returned, so any line on an invoice can be traced back to the call that produced it.