Inference
Call AI models through Jotsu from your own software or a Jotsu application, with server-side authentication, usage tracking, and support for common model tasks.
Inference is Jotsu's standard API for calling AI models. You can use it from your own software or from an application built with Jotsu.
How inference works
Your server sends a model request to Jotsu using an API key. Jotsu authorizes the request, routes it to the model provider, and records the usage for your organization.
Inference supports common model tasks such as classification, extraction, summarization, generation, and tool use. Available models depend on your organization's credit entitlement and the feature making the request.
Use inference from an application
Applications built with Jotsu can call the same inference service. Credentials stay on the server and are not exposed to the browser.
For example, a support application can send a new request to inference and receive a category, urgency, and summary.
Ask Codex to add the server-side call, validate the model response, and handle failures without losing application data.
How inference is charged
Inference first uses your organization's one-time free credit. After that reaches zero, an organization owner must buy AI credits up front. Each request deducts credit based on the selected model's token cost.
See Inference pricing for the complete calculation.
Next steps
- Review the API reference.
- Follow the Add an AI action quickstart.