Models and Usage
Model inference is remote: EngineForge sends the prompt and selected working context through its gateway to a provider enabled by the current service configuration.
Configured providers
The current provider registry supports Anthropic and OpenAI. The model list available to an account can change with service configuration.
Request path
- You submit a task from EngineForge.
- EngineForge assembles the prompt and selected context needed for the request.
- The request goes to the EngineForge gateway.
- The gateway routes it to the configured remote model provider.
- The response streams back into the active Crew session.
Local versus remote
The project index and engine bridge are local. Selected index-derived or file context becomes remote data when included in a model request.
What a request costs you
Each account has one monthly usage allowance, presented as a percentage. Every request spends a share of it, and the size of that share depends on the model: the lineup spans roughly two hundred times between its cheapest and its most expensive model, so the same allowance is many requests on one and few on another.
The allowance resets monthly and does not roll over. Pre-paid top-ups never expire and are spent only after the allowance is exhausted. When both are empty every model is refused, including the cheapest one, because the check is a single allowance comparison that does not consider which model was selected.
Limitations
Provider retention, training use, deletion, and regional processing require approved policy evidence and are not asserted by this page.