One API key for every frontier model.

What happens on every request

The gateway runs the same ordered checks for every call. Nothing reaches an upstream provider until the key, the account, the budget and the model have all cleared.

1. 01

   Key check

   The bearer token is SHA-256 hashed and compared in constant time. Revoked keys stop working immediately.

2. 02

   Account & plan

   Suspended accounts and inactive subscriptions are refused before any upstream call is made.

3. 03

   Allocation

   Remaining tokens are checked against the plan allocation, so an exhausted budget fails fast with 402.

4. 04

   Model authorisation

   The requested model must exist, be enabled, and be permitted on the caller's plan.

5. 05

   Routing

   The model's provider adapter is selected at runtime. Upstream credentials stay server-side.

6. 06

   Accounting

   Prompt and completion tokens, latency, cost and status are written to the usage log per request.

• Chat22
• Reasoning8
• Coding2
• Vision1

Drop-in compatible

Change the base URL and the key. The request and response bodies match the OpenAI Chat Completions shape, streaming included, and there is an Anthropic-dialect endpoint at POST /v1/messages for clients that prefer it.

curl https://bandidoz.biz.id/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer rly_live_YOUR_API_KEY" \
  -d '{
    "model": "relayn-sandbox-chat",
    "messages": [
      { "role": "user", "content": "Explain rate limiting in one sentence." }
    ]
  }'

Base URL for this deployment: https://bandidoz.biz.id/v1. Create an account to mint a key and read the full reference — error codes, rate limits and response headers are documented per endpoint.

Plans

Allocations and rate limits are enforced by the gateway, not just displayed here. The monthly tiers are not billed — switching between them applies immediately and no card is ever charged. Unlimited is the one thing that costs money: a single Rp5.000 QRIS payment that never renews.