Quickstart

Create an account and one is issued during setup. Keys are shown once, so put it straight into your environment:

export TOKENIFY_API_KEY="tk-live-..."

Requests are billed per token against prepaid credit, and a request with no credit behind it is refused with a 402. New accounts get a small amount to try the API with; after that, top up by card or stablecoin. Credit does not expire.

The base URL is https://api.tokenify.dev/v1 and the key goes in an Authorization: Bearer header. Nothing else about your code changes.

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["TOKENIFY_API_KEY"],
    base_url="https://api.tokenify.dev/v1",   # the only line that changes
)

response = client.chat.completions.create(
    model="deepseek/deepseek-v4-flash",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

It is the shape your SDK already expects, with two additions worth knowing about. Cached prompt tokens are reported separately, because they are billed at a fraction of fresh ones, and two headers come back: the canonical model id we resolved your request to, and a request id.

{
  "id": "chatcmpl-...",
  "object": "chat.completion",
  "model": "deepseek/deepseek-v4-flash",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Hello! How can I help?" },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 9,
    "completion_tokens": 8,
    "total_tokens": 17,
    "prompt_tokens_details": { "cached_tokens": 0 },
    "completion_tokens_details": { "reasoning_tokens": 0 }
  }
}

x-tokenify-model: deepseek/deepseek-v4-flash
x-tokenify-request-id: req_aba350ea35840912b70756be

Quote x-tokenify-request-id in a support conversation and we can find the exact request, including what it cost and how it was served. The same id is on the response body as id, prefixed chatcmpl-, so one string covers your logs and ours.

• Streaming — tokens as they are generated, and where the usage figures arrive.
• Prompt caching — how a long system prompt gets much cheaper on the second turn.
• Models and pricing — what each one costs and what it can do.