Quickstart

1. Get an API key

Sign up at app.aisetu.ai (email OTP, or Google/ Microsoft SSO). Creating your first workspace mints a workspace API key — tt_live_… in production, tt_test_… in test mode. You can create additional keys later from the workspace's Settings → Keys tab.

New workspaces start at $0 balance. Visit the top-up link from the dashboard before your first call, or it returns wesence.insufficient_credits.

Export it:

export AI_SETU_API_KEY=tt_live_…

2. Make your first call

Every request goes to https://gateway.aisetu.ai/v1. Pick whichever of these matches your stack — they all hit the same endpoint.

curl

curl https://gateway.aisetu.ai/v1/chat/completions \
  -H "Authorization: Bearer $AI_SETU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4o-mini",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

TypeScript — `@ai-setu/client`

The official SDK — the openai package, subclassed, with no baseURL to set and workspace balance surfaced on every call.

npm i @ai-setu/client

import { AiSetu } from '@ai-setu/client';

const client = new AiSetu(); // reads AI_SETU_API_KEY from env

const res = await client.chat.completions.create({
  model: 'openai/gpt-4o-mini',
  messages: [{ role: 'user', content: 'Hello!' }],
});
console.log(res.choices[0].message.content);
console.log(client.lastBalance); // { micros, isLow, topUpUrl }

TypeScript — stock `openai` SDK

Already using the openai package? Point its baseURL at the gateway and change nothing else:

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: process.env.AI_SETU_API_KEY,
  baseURL: 'https://gateway.aisetu.ai/v1',
});

const res = await client.chat.completions.create({
  model: 'openai/gpt-4o-mini',
  messages: [{ role: 'user', content: 'Hello!' }],
});
console.log(res.choices[0].message.content);

Python — `openai` SDK

There's no dedicated AI Setu Python package yet; the stock openai Python SDK works the same way, pointed at the gateway:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AI_SETU_API_KEY"],
    base_url="https://gateway.aisetu.ai/v1",
)

res = client.chat.completions.create(
    model="openai/gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(res.choices[0].message.content)

3. Pick a model

openai/gpt-4o-mini above is a hard override — that exact provider serves. You can also send a bare model id (e.g. claude-sonnet-4-5) and let AI Setu pick the provider, or route through your own key with @<slug>/<model>. See Models, Routing, and BYOK.

4. Handle errors

Non-2xx responses use the OpenAI error envelope shape plus an AI Setu-specific code. @ai-setu/client exports type guards so you don't have to parse the envelope yourself:

import { isInsufficientCreditsError, getInsufficientCreditsDetails } from '@ai-setu/client';

try {
  await client.chat.completions.create({ model: 'openai/gpt-4o-mini', messages: [...] });
} catch (err) {
  if (isInsufficientCreditsError(err)) {
    const { topUpUrl } = getInsufficientCreditsDetails(err);
    console.error(`Out of credits. Top up at ${topUpUrl}.`);
  }
}

See the full list in Errors.

Routing — how AI Setu picks a provider, and what happens when one fails.
BYOK — route through your own provider keys.
Caching — skip the provider entirely on repeat requests.
Agent Quickstart — wiring up an autonomous agent instead of a human-driven app.