OpenAI proxy

POST /v1/chat/completions — speaks the OpenAI Chat Completions protocol. Mintoken accepts the same request body and returns the same response body. You keep using the OpenAI SDK.

Endpoint

Base URL: https://api.mintoken.in
Path: POST /v1/chat/completions

Basic example

from openai import OpenAI

client = OpenAI(
    base_url="https://api.mintoken.in/v1",
    api_key="mt_live_xxxxx",
    default_headers={"X-Provider-Key": "sk-proj-xxxxx"},
)

response = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[
        {"role": "system", "content": "You are a database expert."},
        {"role": "user", "content": "How does connection pooling work?"},
    ],
    max_tokens=400,
)

print(response.choices[0].message.content)

Request headers

Response headers

Streaming

Pass stream: true exactly as you would with the OpenAI API. Mintoken forwards the SSE chunks as they arrive, adds the X-Mintoken-Intensity response header, and tracks usage after the stream closes.

stream = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Count to 10"}],
    stream=True,
)

for chunk in stream:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="", flush=True)

Overriding intensity per request

Each API key has a default intensity (set at creation or editable in the dashboard). You can override it on a specific request with a header or query param:

curl https://api.mintoken.in/v1/chat/completions \
  -H "Authorization: Bearer mt_live_xxxxx" \
  -H "X-Provider-Key: sk-proj-xxxxx" \
  -H "X-Mintoken-Intensity: ultra" \
  -H "Content-Type: application/json" \
  -d '{ "model": "gpt-4o-mini", "messages": [...] }'

If you want mintoken to pick automatically, enable smart detection on the key — see Smart detection.

Supported models

Any model your OpenAI account has access to. Mintoken doesn't maintain an allowlist — the model string is forwarded as-is. At time of writing, the common picks:

• gpt-4o-mini — best cost/quality for most workloads
• gpt-4o — highest quality, higher cost
• gpt-4.1-mini / gpt-5-mini — newer, check your account access
• o1 / o3 — reasoning models; mintoken compression applies to the visible output, not the reasoning traces

Error responses

Upstream OpenAI errors (4xx and 5xx) are relayed back to you with the original status code and body. Mintoken-specific errors use the following status codes: