A Claude Code skill that saves 30-60% on costs through concise responses (40-70% output token reduction), smart model routing, and efficient workflow patterns.
Claude Code (official plugin):
/plugin marketplace add Sagargupta16/claude-cost-optimizer
/plugin install cost-mode@sagargupta16-claude-cost-optimizerMulti-agent (Cursor, Cline, Codex, etc.):
npx skills add Sagargupta16/claude-cost-optimizerOnce installed, activate with:
/cost-mode # Toggle on (standard intensity)
/cost-mode lite # Professional brevity, full sentences
/cost-mode standard # Concise fragments, skip filler (default)
/cost-mode strict # Telegraphic, max savings
/cost-mode off # Resume normal behavior
- Cuts filler: No pleasantries, hedging, restating questions, or trailing summaries
- Suggests cheaper models: Recommends Haiku for simple tasks, Sonnet for standard work
- Suggests CLI tools: Points to
prettier,eslint --fix,gitinstead of burning LLM tokens on deterministic tasks - Session awareness: Reminds you to
/compactafter 20+ turns, start fresh sessions for new tasks - Minimal code gen: Diffs over rewrites, no obvious comments, no speculative error handling
- Full technical accuracy
- All code blocks unchanged
- Security warnings at full clarity
- Destructive operation confirmations
- Detailed explanations when you explicitly ask
| Intensity | Output Token Reduction | Estimated Cost Savings | Best For |
|---|---|---|---|
| lite | 20-40% | ~10-20% | Team-visible work, PRs, shared sessions |
| standard | 40-60% | ~20-35% | Daily development, solo coding |
| strict | 60-70% | ~30-45% | High-volume sessions, budget-constrained |
Output tokens cost 5x more than input tokens, so even moderate output reduction has significant cost impact. Combined with model routing and CLI suggestions, total savings reach 30-60%.
MIT