Verified:GitHub

ecomode

Route LLM subagents to lower-tier models with escalation to Sonnet and Opus.

Updated Feb 20, 2026

Search Tags:#claude-code #token-optimization #model-selection #agent-delegation #autonomous-execution #llm-routing #cost-control

Ready to run

npx skills add https://github.com/ssccio/cc-on-web-base --skill ecomode-ssccio

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: ecomode
Source: https://github.com/ssccio/cc-on-web-base/tree/main/skills/ecomode
Command: npx skills add https://github.com/ssccio/cc-on-web-base --skill ecomode-ssccio

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It reduces token cost by routing work to lower-tier models while still ensuring successful execution through controlled escalation.

Core Features & Use Cases

Token-efficient model routing modifier: Prefer LOW (Haiku) by default, use MEDIUM (Sonnet) when complexity warrants, and only escalate to HIGH (Opus) if essential.
Combines with existing execution modes: Works as a modifier alongside modes like ralph (persistence), ultrawork (parallel execution), and autopilot (full autonomy).
Cost-optimized agent delegation rules: Directs common task types (code changes, analysis, search, documentation) to specific low/medium/high agent tiers with matching model choices.

Quick Start

Ask the system to run your task with Ecomode enabled so it prefers Haiku routing and only escalates to Sonnet or Opus when necessary.

Frequently Asked Questions about ecomode

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I route Claude Code subagents to cheaper models for token optimization?▼

Token optimization routes LLM subagents to lower-tier models like Haiku by default, escalating to Sonnet or Opus only when complexity requires, reducing cost without breaking execution flow.

What is the best way to control LLM routing costs for autonomous coding agents?▼

LLM routing cost control applies tiered model selection, directing common tasks like code changes and search to Haiku, while escalating to Sonnet on failure or Opus for essential planning.

Can I use cost-optimized model selection with parallel execution and autonomous workflows?▼

Cost-optimized model selection works as a modifier alongside execution modes like parallel execution and full autonomy, applying tiered routing to delegated agent workflows without interrupting them.

When should I escalate from Haiku to Sonnet or Opus during agent delegation?▼

Agent delegation escalates from Haiku to Sonnet upon task failure, reserving Opus exclusively for essential planning or critique, ensuring model quality matches complexity only when required.

Does token-efficient model routing work for analysis and documentation generation tasks?▼

Token-efficient model routing supports analysis, search, and documentation generation by directing these common task types to specific low or medium agent tiers with matching model choices.

Why does routing all tasks to a single cheap model break autonomous execution?▼

Routing all tasks to a single model lacks escalation, causing execution failures on complex planning; tiered routing solves this by defaulting to low-tier models and escalating only on failure.