Skip to content

The best models for your coding agent, for less.

Prepay a USD Balance, get one API Key, and call every model in the Catalog from Claude Code, Codex, Cursor or OpenCode. You pay below OpenRouter’s prices, per Token, and your Balance never expires.

  • Claude Code
  • Codex
  • Cursor
  • OpenCode
  • Cline
  • Aider
Today’s prices per 1M Tokens
  • GPT-6 Luna99% cheaper
    OpenRouter
    $0.10 in · $0.50 out
    You pay
    $0.001 in · $0.005 out

Input + output, live from the Catalog.

Why it costs less

Every model shows two prices: the OpenRouter Price for the same model, and our Sell Price. The difference is yours.

Sell Price below OpenRouter Price
We set each Sell Price below what OpenRouter charges for the same model. The Pricing page and every Request show what you saved.
Cache reads are cheaper still
Coding agents resend the same context all day. Cached input Tokens are billed at a fraction of the input price.
Pay only for Tokens you use
No subscription and no monthly minimum. A Request deducts exactly Tokens × Sell Price from your Balance.

Set up in two minutes

Log in from your terminal once, then let the CLI write your tool’s config. Or copy a ready snippet from the Setup page.

  1. 1Create an account and verify your email.
  2. 2Run the login command and approve it in your browser.
  3. 3Run setup for your tool and start coding.
Read the quick start
Terminal
npx centrouter login
npx centrouter setup claude    # or codex, opencode
claude

Top up from small amounts

Prices are in USD with a Rupiah estimate next to them. Online payment is not open yet: for now you add Balance with a Voucher code on the Balance page.

Today, $1 ≈ Rp 18.200.

From developers who use it

In their own words, quoted with their permission.

  • “keren murah banget dibanding yang lain”

    Rahmat S · Backend Developer, Jakarta
  • “easy to use and cheapest ai gateway in the world”

    Michael K · Project Manager, New York

Your first API Key is ready the moment you sign up.

Check the live status of every model before you start.