From OpenRouter

OpenRouter and Cloakray both speak the OpenAI format, so most apps move by changing the base URL, the key and the model IDs.

Base URL and key

 const client = new OpenAI({
-  baseURL: "https://openrouter.ai/api/v1",
-  apiKey: process.env.OPENROUTER_API_KEY,
+  baseURL: "https://api.cloakray.com/v1",
+  apiKey: process.env.CLOAKRAY_API_KEY,
 });

Model IDs

OpenRouter prefixes models with their maker; Cloakray IDs have no prefix, and use the maker's own spelling.

OpenRouter	Cloakray
`openai/gpt-5.5`	`gpt-5.5`
`anthropic/claude-sonnet-4.6`	`claude-sonnet-4-6`
`google/gemini-3.6-flash`	`gemini-3.6-flash`
`deepseek/deepseek-v4-flash`	`deepseek-v4-flash`

Look up each model you use on the models page. Variant suffixes such as :free, :nitro or :online are not supported.

Request options

OpenRouter option	On Cloakray
`reasoning: { effort }`	`reasoning_effort` on Chat Completions — see reasoning.
`provider` preferences, `models` fallback list, `route`	Not needed. You choose the model; Cloakray serves it. Drop these fields.
`transforms`, `plugins`	Not supported. For web results, use the search API.
`usage: { include: true }`	Not needed. Every reply's `usage` includes `cost`, the amount charged — see usage and cost.
`HTTP-Referer`, `X-Title` headers	You can leave them in; they have no effect.

Standard parameters — tools, response_format, stream, max_tokens, temperature — work as before.

Account endpoints

GET /v1/key returns the key's label, usage and limit in the same shape as OpenRouter's, so scripts that read it keep working. GET /v1/credits returns balance and total_used. See authentication.

Pricing

You pay the model's list price minus any discount on your account, with no fee on top-ups or requests.