AI Gateway
Call Ertas models from your app through one OpenAI-compatible API. One base URL, one key, and per-token prices paid from your Ertas credits.
On this page
The Ertas AI Gateway lets your app, agent or script call models hosted by Ertas through the OpenAI API format. If your code already uses an OpenAI SDK, you change two settings, the base URL and the API key, and pick an Ertas model ID.
| Setting | Value |
|---|---|
| Base URL | https://gateway.ertas.ai/v1 |
| API key | An Ertas key starting etk_, created in Gateway → API Keys |
| Model | An Ertas model ID, such as ertas-ai/medertas-27b |
Every API call is priced per token at the model's price and paid from the same credit balance you use for training. US$1 buys 100 credits.
What you can call today
- MedErtas, our open-weight medical models:
ertas-ai/medertas-31bandertas-ai/medertas-27b. See Models and prices. - Chat Completions with streaming, tool calls and structured output. See Chat Completions.
- The Responses API for stateless calls, with streaming and structured output. Stored responses are coming soon. See Responses API.
Coming soon: deploying the models you train in Model Studio on the gateway, and bringing your own provider keys.
First API call
Read the first api call guide.
API keys
Read the api keys guide.
Activating your endpoint
Read the activating your endpoint guide.
Models and prices
Read the models and prices guide.
Chat Completions
Read the chat completions guide.
Responses API
Read the responses api guide.
Coding Agents
Read the coding agents guide.
Switch from another gateway
Read the switch from another gateway guide.
API credits and usage
Read the api credits and usage guide.
Errors and retries
Read the errors and retries guide.
Data And Privacy
Read the data and privacy guide.
Limits
Read the limits guide.
How it fits with the rest of Ertas
- Run models on the gateway, starting today.
- Train your own models on your data in Model Studio.
- Improve a model continually from your traffic on the adaptive endpoint, by invitation.