APIKO
OpenAI-compatible · billed on real usage

One API for every AI model

Apiko routes your OpenAI-compatible requests across providers with automatic retry and failover — billed by real token usage, refunded on failure.

  • Live per-token pricing
  • Failed calls auto-refunded
  • Free trial credit

no credit card required

Live pricing
4 modelsView all
Models from
Anthropic
OpenAI
Google
DeepSeek
26+ models
01 — Quickstart

Ship in three steps

If your code speaks OpenAI, it already speaks Apiko.

  1. 01

    Create an API key

    Sign up and mint a key from the console. The full key is shown once and never stored in plaintext.

    Create account
  2. 02

    Point your base URL

    Change one line in your existing OpenAI SDK setup. No new SDK to learn.

    base_url = "https://api.apiko.example/v1"
  3. 03

    Call any model

    Use one key for every model behind the gateway — usage and spend land in one dashboard.

    See the catalog
curl https://api.apiko.example/v1/chat/completions \
  -H "Authorization: Bearer $APIKO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

example host shown — your deployment's gateway address is documented in the docs

02 — Agent-ready

Let your coding agent wire it up

Paste one line into Claude Code, Cursor, or any agent — it reads our llms.txt and configures everything.

Read https://apiko.app/llms.txt and configure this project to call AI models through it. Verify with a test request to gpt-5.2.

llms.txt →
03 — Models

Models with transparent pricing

Live prices from the gateway — what you see is what gets billed.

View all models
04 — Billing

billing guarantees — how every call settles

Cost of failed calls
$0

Upstream failure? The pre-held quota is automatically refunded.

Billing basis
real usage

Settled per request from the provider's actual token counts.

Wallet
one balance

A single balance across every model and provider.

05 — Infrastructure

Built like infrastructure, not a wrapper

A weighted channel pool with health-aware routing sits between you and the providers.

Automatic retry & failover

Requests reroute across a weighted channel pool when a provider fails.

Health-aware routing

Failing credentials are disabled instantly — traffic only hits healthy channels.

Pre-hold, settle, refund

Billing pre-holds on request, settles on real usage, refunds on failure.

Per-request logs

Tokens, latency, and exact cost recorded for every call.

OpenAI-compatible

Official SDKs and any HTTP client work unchanged.

One balance

A single balance across every model and provider.

> retry: automatic, laddered across channel priorities
> failover: weighted channel pool per model
> auto_ban: credential failures disable a channel instantly
> billing: pre-hold → settle on real usage → refund on failure
gateway behavior — verified end to end

Frequently asked questions

Start calling models in minutes

Create a key, change one line, and route your first request through the gateway.

free to sign up · no credit card