ecomode
Low-token, low-cost execution mode — routes tasks to cheaper models and prunes context aggressively to reduce spend.
ecomode switches the active session to a cost-optimised profile. It re-routes all subsequent agent dispatches to lower-tier models, enables aggressive context pruning, and disables high-context features like deep research passes. Use it when you want results at minimal token cost and can accept a modest quality trade-off.
Purpose
Not every task needs opus-tier reasoning or a full context window. ecomode makes the trade-off explicit: you accept slightly lower fidelity in exchange for significantly lower spend. It is particularly useful for repetitive tasks, bulk file operations, or exploratory work where you expect to iterate anyway.
How to invoke
Natural language triggers: "ecomode", "eco mode", "low cost", "cheap mode", "save tokens".
Explicit slash: $ecomode
codex
> ecomodecodex
> enable ecomode for this sessionWhat it does
- Sets the default model tier to haiku or the cheapest available model
- Enables aggressive context pruning: removes prior conversation turns beyond a short window
- Disables high-cost features: deep research, consensus planning, and multi-reviewer passes
- Applies through the rest of the session or until you explicitly disable it
- Reports the estimated cost impact when switching modes
Outputs
- Confirmation that ecomode is active with the current model tier shown
- Updated HUD indicator reflecting the reduced cost profile
- No source code changes