economy-mode

You can reduce AI agent costs by enabling economy mode to switch Layer 3 model selection to lower-cost alternatives. This minimizes expenses across tasks like architecture planning and code generation while preserving your original user intent.

Cost-aware model selection works by applying an economy table to Layer 3 decisions when a runtime toggle is active. It persists settings via a configuration file or session scope, using lower-cost models while respecting Layer 0 overrides and explicit directives.

Yes, you can use economy mode without overriding default model constraints. The feature keeps user-defined constraints like defaultModel and agentModelOverrides intact while applying the economy table exclusively to Layer 3 auto-selection decisions.

The best way to persist cost optimization settings is to write the economyMode toggle into your config.json file. This applies the economy table to Layer 3 model choices across all future sessions until manually disabled.

Economy mode does not affect triage and code generation tasks differently; it uniformly shifts their Layer 3 model selection to lower-cost alternatives. It minimizes expenses across all applicable tasks without altering the underlying user intent.

You should not use economy mode for complex architecture planning if maximum model capability is required. While it preserves user intent, shifting to lower-cost alternatives inherently minimizes model capability, which may impact highly complex reasoning tasks.