Base URL and authentication

Base URL

Each API format uses a slightly different address; the difference is whether it ends in /v1:

API format	Address	Used by
OpenAI-compatible	https://api.rabah.ai/v1	Most apps and SDKs, Codex CLI, OpenCode
Anthropic	https://api.rabah.ai	Claude Code and other Claude clients
Gemini native	https://api.rabah.ai	Gemini CLI, Google’s official SDKs

No /v1 on the Anthropic address

Claude clients append /v1/messages themselves. If you enter https://api.rabah.ai/v1, the request goes to /v1/v1/messages and fails.

Apps name this setting differently — Base URL, API URL, API Endpoint — but it always takes the address above.

https://hub.rabah.ai also accepts API calls, but use api.rabah.ai: the former is the web console, the latter is meant for API traffic.

Passing the key

Put the key in a request header, either of:

Authorization: Bearer YOUR_KEY

x-api-key: YOUR_KEY

Most OpenAI-compatible apps use the first; Claude clients usually use the second. The Gemini native API also accepts an x-goog-api-key header.

Never put the key in the URL

?key=... or ?api_key=... is rejected with a 400 error. URLs end up in logs, browser history and proxies, where a key leaks easily.

Endpoints

Append these paths to the OpenAI-compatible address https://api.rabah.ai/v1:

Path	Purpose
POST /chat/completions	Chat — the most widely supported endpoint
POST /responses	OpenAI Responses API, used by Codex
POST /messages	Anthropic Messages API, used by Claude clients
GET /models	List the models this key can use
POST /embeddings	Text embeddings
POST /images/generations	Generate images
POST /images/edits	Edit images

You rarely need to build these yourself: set the base URL in your app and it adds the path.

Models and lines

Set model in the request to a model name from the Model Plaza.

A model may be offered on several lines, and one key can use all of them:

• Model name only, e.g. claude-sonnet-4-6: uses the model’s default line.
• model:line-code, e.g. claude-sonnet-4-6:pro: uses that specific line.

Line codes are listed in the Model Plaza. GET /v1/models also lists, for models on several lines, the model:line-code form of each. For how the default line is chosen and what happens when a line is unavailable, see Lines.

Limiting what a key can do

When you create or edit a key you can keep separate keys for separate uses and set:

• Lines & Models: only the lines and models you tick.
• Quota: the most this key can ever spend.
• Rate limits: the most it can spend in 5 hours, 1 day and 7 days.
• Expiry: the key stops working after this date.
• IP restriction: allow or block specific IPs.

For example, give a shared automation script its own key with a quota: if the key leaks, the damage is capped.