Claude, GPT and Gemini APIs, for less.

One OpenAI- and Anthropic-compatible Base URL. Pay per token, see every call in your usage log, and keep the code you already have.

Base URLhttps://api.airoutly.com/v1

Prepaid balance via Stripe.

curl https://api.airoutly.com/v1/chat/completions \
  -H "Authorization: Bearer $AIROUTLY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-5",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

The highlighted line is the only change to your code.

Three steps to your first call

  1. 1

    Create a key

    Sign up with email or GitHub, top up any balance, and create a key with its own spend limit.

    sk-••••••••••••a91f
  2. 2

    Change one line

    Swap the Base URL in your SDK or tool. Model names stay the official ones.

    api.openai.com/v1
    api.airoutly.com/v1
  3. 3

    Pay per token

    Every request shows up in your usage log with tokens and cost, so you can reconcile to the cent.

    200 sonnet-4-5 · 1,284 tok · $0.0119

What you can hold us to

No inflated numbers. These are the things you can verify yourself.

Public prices
Every model's price is listed before you sign up, and it's what you're charged.
A log line for every call
Model, tokens, cost and time for each request.
Card payments through Stripe
We never see or store your card number.
Native protocols
Anthropic Messages, OpenAI Chat and Responses, Gemini, passed through as-is.

Questions developers ask first

Something else? Contact support

Is this the real Claude / GPT / Gemini?
Yes. Requests are routed to the upstream models through our provider partners, and the model name you send is the model that answers.
How is usage billed?
You top up a USD balance. Each request deducts input, output and cache tokens at the per-model price shown on the pricing page. Your usage log shows the exact amount per request.
Does it work with Claude Code and Cursor?
Yes. Set the Base URL and your Airoutly key in the tool's settings. The docs have a step-by-step page for each client.
Are streaming and long requests supported?
Streaming is supported on every model, and recommended. Non-streaming requests that take longer than about two minutes can time out; use stream: true for long outputs.
Do you store my prompts?
We log request metadata (model, tokens, cost, time) for billing. How prompt content is handled by us and our upstream providers is described in the privacy policy.

Your key is two minutes away.

Sign up, top up, paste one Base URL. Keep building.