Claude Code Token-Saver Kit — cut your AI coding bill without writing worse code

---

## Product name

Claude Code Token-Saver Kit — cut your AI coding bill without writing worse code

## Subtitle / tagline

18 battle-tested tactics + drop-in configs that stop the token waste in Claude Code and the Anthropic API. One-time. No subscription.

Heavy Claude Code user watching your token bill climb? Most of that spend is pure waste — re-reading files, using the priciest model for trivial edits, blowing context every turn, and leaving 50–90% discounts switched off. This kit fixes all of it: a ranked playbook, a ready-to-merge cost-control `settings.json`, a model-routing decision table, a drop-in rules file, and weekly spend-report scripts. Quality stays the same. The waste goes away.

## Full description

**The problem.** Claude Code is the best money you spend — right up until the bill makes you flinch. The dirty secret is that a big chunk of that bill buys you nothing: a bloated context re-sent on every turn, Opus doing variable renames, prompt caching left off, batch jobs run at full price.

**What you get.**

- 📘 **The Playbook (`GUIDE.md`)** — 18 tactics ranked by *dollars saved per minute of effort*. The first five are ~80% of the win and take 10 minutes.

- ⚙️ **Cost-control `settings.json`** — ready to merge: sane default model, capped thinking budget, a context-hygiene reminder hook.

- 🧭 **Model-routing table** — exactly when Haiku vs Sonnet vs Opus, so you stop overpaying for easy work.

- 📐 **Drop-in rules file** — nudges the agent toward cheaper behavior automatically, every session.

- 📊 **Spend-report scripts** (Windows + macOS/Linux) — watch the number go down each week.

**Who it's for.** Daily Claude Code users, Anthropic API builders, and teams who want a cost-aware setup as a standard.

**What it's NOT.** Not a hack, not a way to make Claude dumber to save pennies, and it can't change Anthropic's prices. It kills *waste*, not quality.

## FAQ

- **Will this lower my output quality?** No. Every tactic preserves quality; several (plan-first, context hygiene) tend to improve it.

- **I only use Claude Code, not the API — still useful?** Yes. The model-routing, context-hygiene, `/clear`, and thinking-cap tactics are all Claude Code-side.

- **Do prices/model names go out of date?** The numbers can move; the kit tells you to confirm current pricing in Anthropic's docs. The *tactics* stay valid.

- **Refunds?** Yes — 14 days, no friction.

## Guarantee

If it doesn't help, reply to your receipt within 14 days for a full refund.