aicentos

Posted on Nov 12 Originally published at Medium

aicentos: One API Key for 94 LLM Models — Quickstart in 3 Steps

#ai #llm #api #tutorial

TL;DR

aicentos is an enterprise LLM API gateway: 94 models from 13 providers — OpenAI, Anthropic, Google, xAI, DeepSeek and more — behind one key and one endpoint that speaks OpenAI, Anthropic and Gemini protocols. Sign up, create a key, swap the base URL. Your SDKs and code stay the same.

Docs: https://docs.aicentos.com
Live status: https://status.aicentos.com
GitHub: https://github.com/aicentos-gateway

The problem: one key per provider, one bill per model

If you use Claude Code, Codex, Cursor, or any OpenAI-compatible client, you know the drill: a different API key for every provider, a different base URL for every tool, a different billing page for every model — and a different failure mode when any one of them goes down.

A gateway collapses that into a single integration point. You point your existing tools at it, and it routes to the model you asked for.

What aicentos actually does

Three things matter in practice:

1. Failover is built in

Upstream is a self-built pool of official accounts, all connected directly to the vendors' official APIs. If one account fails on a request, the gateway automatically retries the next account serving the same model. You change nothing in your code.

2. Your prompts stay yours

Usage logs record metadata only — model, token counts, latency. Request and response bodies are not stored.

3. Keys are governable

Every key can carry its own spending cap, model allowlist, IP allowlist and expiry, and can be revoked at any time. Create one key per project, and a leak stays a small problem.

Billing is pay-as-you-go and the balance never expires. Availability and latency are public on the status page, and enterprise plans cover contracts, invoicing and self-hosted deployment.

Quickstart

Step 1 — Sign up

Go to aicentos.com and create an account. New accounts get ¥1 of free credit, so you can make real calls before topping up.

Step 2 — Create an API key

In the console, open API Keys and click Create API Key. Keys look like sk-xxxxxxxxxxxx and are shown in full only once, at creation — copy it immediately and store it somewhere safe.

Never hard-code a key into frontend code or commit it to a public repo.

Step 3 — Point your client at the right Base URL

This is the step people get wrong, because the address depends on the protocol:

Scenario	Base URL
Codex / OpenAI-compatible format	`https://aicentos.com/v1`
Claude Code / Anthropic format, and other clients	`https://aicentos.com`

For Anthropic, use the root address only — do not append /v1/messages. The client adds the path itself. You can also check and copy the address from the API address dropdown on the console's API Keys page.

Verify it works

export AICENTOS_API_KEY="sk-xxxxxxxxxxxx"

curl https://aicentos.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $AICENTOS_API_KEY" \
  -d '{
    "model": "claude-sonnet-5",
    "messages": [
      { "role": "user", "content": "Hello, aicentos!" }
    ]
  }'

Or just list the models — any JSON response means your key and base URL are good:

curl https://aicentos.com/v1/models \
  -H "Authorization: Bearer $AICENTOS_API_KEY"

The same key works from the OpenAI SDKs. In Python:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AICENTOS_API_KEY"],
    base_url="https://aicentos.com/v1",
)

resp = client.chat.completions.create(
    model="claude-sonnet-5",
    messages=[{"role": "user", "content": "Introduce yourself in one sentence."}],
)
print(resp.choices[0].message.content)

69 client guides

There are step-by-step illustrated guides for 69 AI coding tools and chat clients — Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Cline, Roo Code, Cherry Studio and more. Fill in the key, save, done:

👉 docs.aicentos.com/clients

FAQ

I get a 401 — what do I check first?
Your key, and your balance. A wrong, expired or revoked key returns 401, and so does an empty balance. Recreate the key in the console if in doubt, and check the balance in Wallet.

Why does Claude Code fail when the same key works in Codex?
Almost always the Base URL. Codex / OpenAI-compatible clients use https://aicentos.com/v1; Claude Code / Anthropic uses the root https://aicentos.com with nothing appended.

Do I have to top up before trying?
No. New accounts get ¥1 of free credit, and the balance never expires. Top-ups start at ¥1, with discounts from ¥100 up.

Where do I see what I spent?
The console's Dashboard shows balance, total usage, request counts and per-model trends.

DEV Community