Advanced

Model Routing

Let the platform select a model while preserving budget and policy constraints.

Auto Model Selection

Model routing can select a model for a task based on capability, context window, cost cap, provider quality, and required protocol.

The response remains auditable: requested model, actual model, actual provider, and route reason are returned together.

Agent presets keep long-running coding sessions sticky to cache-friendly providers unless reliability or policy requires a change.

This reduces cache misses and makes token spend more predictable for tool-heavy workflows.