Core Concepts
New here? This page explains the terms you’ll see, in everyday language. No technical background required.
What is Nexotao?
Nexotao is a bridge for using the best AI models (like Claude, GPT, and DeepSeek) while paying in Rupiah via QRIS — no foreign credit card, no monthly subscription. You top up a balance, then use it as needed.
AI model
A model is the “brain” that answers your questions. Each model has a different character:
- Some are cheap and fast — good for simple, high-volume tasks (e.g. sorting thousands of comments).
- Some are smart and powerful — good for heavy work like writing code or analyzing long documents.
Analogy: like choosing a vehicle. A scooter for short, cheap trips; a truck for heavy loads. You pick by the job. See the options in Models & Pricing.
Token
AI doesn’t count “words” — it counts tokens, small chunks of text. One word is usually 1–2 tokens. Roughly 1 token ≈ 4 characters.
Why it matters: cost is per token. The longer your question and answer, the more tokens, the higher the cost. It’s also why AI answers are sometimes length-limited.
Example: “Hello, how are you?” is about 5 tokens.
API key
An API is how a program (code) talks to Nexotao — different from opening a website
in a browser. An API key is your private “key”, shaped like sk-nexo-..., that
proves a request comes from your account (and bills your balance).
Treat it like a password: don’t share it, don’t put it in publicly visible code. If it leaks, revoke it in the dashboard and make a new one.
Details in Authentication.
Endpoint
An endpoint is the “address” your request goes to, e.g.
https://api.nexotao.com/v1/chat/completions. Nexotao has a few endpoints for
different formats. Picking the right one matters — guide at
Endpoints & Formats.
Balance & cost
You top up a Rupiah balance (from Rp 10,000 via QRIS), then each request deducts
by the tokens used. No subscription; the balance doesn’t expire. Every response
tells you the exact cost via the X-Cost-Rp header. Details in
Billing & Pricing.
Context window
Each model can only “remember” so many tokens at once in a conversation — that’s its context window. If a conversation gets too long, the oldest part can be “forgotten” or the request rejected. How to handle it: Context Window.
Streaming
Streaming means the answer appears bit by bit (like typing), instead of all at once when finished. Useful so long answers feel fast.
Got the basics? Continue to the Quickstart for your first request, or try it directly in the Playground — no code required.