Muse Wonder API

Use OpenAI-compatible chat completions with the Muse Wonder model.

Create an API key

Sign in, add credits, and create a key in API key settings. Send the key in the Authorization header as Bearer YOUR_API_KEY. Keep it on your server.

Send a conversation

POST /api/v1/chat/completions

{
  "model": "muse-wonder",
  "messages": [
    {
      "role": "user",
      "content": "Explain a JavaScript closure."
    }
  ],
  "max_tokens": 512
}

Set stream to true for server-sent events. GET /api/v1/models lists supported models. Messages use system, user, assistant and tool roles. Function tools are supported; the caller executes requested functions and sends their results back.

Credits and limits

API rates are $ 32 per million input tokens and $ 160 per million output tokens, converted to credits at $0.01 per credit and rounded up to a minimum of one credit. These are this service’s rates, separate from upstream provider rates. Reasoning tokens count as output usage. Cached input is conservatively billed at the regular input rate.

Requests accept up to 48,000 estimated input tokens and 8,192 output tokens. Credits are reserved before generation and settled against actual provider usage. Failed requests release the reservation. API-key rate and spending limits also apply.

Recover from errors

A 401 response means the key could not be authenticated. A 402 means the balance cannot cover the request. A 429 means a rate or spending limit was reached. Retry temporary failures with backoff. Use an Idempotency-Key header to avoid duplicate billing when retrying the same request.

Legacy typed-decision endpoints are unavailable. Use chat completions for text and coding responses.