A small gateway with a simple job
MagicAI gives developers one OpenAI-compatible endpoint and one balance for multiple AI models. You keep your SDK, pick a model by name, and pay only for the tokens you use.
How it works
Your request arrives at magicai.lol/v1, is checked against your key and balance, and is forwarded to an upstream endpoint serving that model. The response streams back to you unchanged in OpenAI format, and the token counts are used to bill your credit. Failed upstream attempts are retried before an error is returned.
What we believe
Prices are listed per token. Speed stats come from our own logs, with sample sizes.
Models are either Live or Coming soon. We don't list anything as available that isn't.
Standard API format means you can leave as easily as you arrived.
Where we are
MagicAI is a young service. The set of live models is growing; catalog entries marked Coming soon are on our list but not callable yet. See Status for measured performance and Contact to request a model.