NoTrack API · api.notrack.ai

The uncensored model, as an API

The same model that runs notrack.ai, on our own hardware. Point any OpenAI SDK at us: change the base URL and the key, nothing else. No logs, no system prompt of ours, no topic filter.

Get an API key Read the docs

Pricing

$0.25 / 1M

per 1M input tokens — everything you send: system prompt, history, the new message

$1.00 / 1M

per 1M output tokens — what the model writes

free credit on your first key, so you can integrate and test before paying

Prepaid credit, no subscription, nothing renews by itself. One model, 32K context. Refused or failed requests are not billed.

Three lines to switch

Base URL https://api.notrack.ai/v1, model notrack-uncensored. Streaming, system prompts and the usual parameters work as they do everywhere else.

curl https://api.notrack.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-notrack-…" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "notrack-uncensored",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

# pip install openai
from openai import OpenAI

client = OpenAI(base_url="https://api.notrack.ai/v1", api_key="sk-notrack-…")
r = client.chat.completions.create(
    model="notrack-uncensored",
    messages=[{"role": "user", "content": "Hello"}],
)
print(r.choices[0].message.content)

// npm install openai
import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://api.notrack.ai/v1", apiKey: "sk-notrack-…" });
const r = await client.chat.completions.create({
  model: "notrack-uncensored",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(r.choices[0].message.content);

# SillyTavern → API Connections
API:                 Chat Completion
Chat Completion Source: Custom (OpenAI-compatible)
Custom Endpoint:     https://api.notrack.ai/v1
Custom API Key:      sk-notrack-…
Model:               notrack-uncensored     # press "Connect", then pick it

What you get — and what you do not

Your prompt is the whole prompt

We prepend nothing and run no topic filter. Adult fiction, dark themes, strong language: the model answers as written.

Nothing written to disk

Prompts and replies are never stored — not by the gateway, not by the model servers. We keep a request id and token counts, because that is the bill.

Our own model, our own hardware

A companion tune you will not find on aggregators, served from GPUs we rent and control. No third-party provider sees your traffic.

OpenAI-compatible

The official SDKs, SillyTavern, LangChain, Open WebUI, curl — anything with a custom base URL works unchanged, streaming included.

Limits you can plan around

4 parallel requests and 120 requests per minute per key, 32K context. A daily spending ceiling per key, so a leaked key cannot drain the balance.

One rule that never switches off

Anything sexual involving a minor is refused, in code, on every request. It is the only content rule we enforce.

How it works

Create an account and a key

A nickname and a password — nothing about you. The key is shown once; give each app its own.

Test on the free credit

$1 arrives with your first key. Enough to wire up a client and run an evening of conversation.

Top up when you need more

Card or USDT/USDC, from $5. The balance shows on every response header, so your code can see it coming.

Paying

Prepaid packs, credit never expires, no fee on top. Card (Stripe) or crypto (USDT on TRON, USDC on Base). Unused credit can be refunded within 14 days.

$5.00$5.00 credit

$10.00$10.00 credit

$25.00$25.00 credit

$50.00$52.50 credit · +5% bonus

$100.00$110.00 credit · +10% bonus

Questions

Is it really uncensored?

There is no system prompt of ours and no topic filter. The single exception is enforced in code and cannot be turned off: sexual content involving a minor is refused. Everything else is between you and the model.

What does a typical roleplay turn cost?

At $0.25 per 1M input and $1.00 per 1M output tokens, a turn with a 10K-token history and a 300-token reply costs about a third of a cent. One dollar is roughly 250–300 such turns.

Do you keep my prompts?

No. Prompts and replies are not written to disk. We keep the request id, the key id and the token counts, because that is what the bill is made of. Safety refusals are logged by category, without the text.

Which model is it?

notrack-uncensored — our own companion tune of an open 35B mixture-of-experts model, the same one behind the chat on notrack.ai, running on GPUs we control. There is one model, so there is nothing to choose and nothing to route.

Does credit expire? Is there a subscription?

Credit never expires and there is no subscription: nothing renews and nothing is charged by itself. Top up when you want, by card or in USDT/USDC. Unused credit can be refunded within 14 days of purchase.

How big is the context? What are the limits?

32K tokens for prompt plus completion. Per key: a handful of parallel requests and a per-minute request limit, both returned in the response headers; and an optional daily spending ceiling you set yourself.

Get an API key

Everything the gateway does, in one page: Read the docs · Acceptable Use