claudeers.
// Claude Plugins

cache-maxxing

A bar above your chat to see when Claude's prompt cache expires, and what your next message will cost, before you send it.

Install with your AI

Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.

Install and set up cache-maxxing (claude-plugin project) into my current project.
Found on https://claudeers.com/cache-maxxing
Repo: https://github.com/5m0k3/cache-maxxing
Homepage/docs: —
Detected install method: claude-plugin → /plugin install cache-maxxing@5m0k3/cache-maxxing
Category: plugins. Platforms: cli, api, desktop.
Read the repo's README for exact setup and env vars, then install it and wire it into my project.

Claudeers Health Verdict:
unknown; community-verified: false. Confirm the source before running anything.
// or install directly (claude-plugin)

⚠ Unverified / not recently updated — review before pasting a run-this config.

/plugin marketplace add 5m0k3/cache-maxxing
/plugin install cache-maxxing@5m0k3/cache-maxxing
// or clone
git clone https://github.com/5m0k3/cache-maxxing

// compatibility

Platformscli, api, desktop
Operating systems—
AI compatibilityclaude
LicenseMIT
Pricingopen-source
LanguageTypeScript

Get your FREE $2.50 API credits to access TickAtlas financial data ↗
cache-maxxing icon

cache-maxxing

See when Claude's prompt cache expires, and what your next message will cost, before you send it.


About

Ever feel like your Claude subscription quota disappears way faster than it should? Usually, it's not how much you're building - it's when you hit Send.

Claude keeps your chat "warm" only for a while so follow-up messages barely dent your limit. But wait too long between messages, and Claude has to re-read your entire chat history at full cost - up to 80× more quota than a normal reply.

cache-maxxing puts an easy-to-read bar right above your chat box (in the desktop app, VS Code, and terminal) to protect your quota while you build:

  • Know before you hit Send: See a live countdown timer and whether your next message is cheap (🟢), running out of time (🟡), or about to take a huge bite out of your limit (🔴).
  • Never lose your flow with /cmaxx-handoff: When a long chat gets too expensive to continue, run one command to hand off a smart summary of your progress into a fresh chat—no re-explaining your project, and no massive quota spike.
  • Zero setup headaches, zero slowdown: Everything is calculated right on your machine using numbers Claude Code already tracks. The bar sends nothing anywhere.
Green bar: cache active, next message cheapYellow bar: cache ending soon
🟢 Cache active. Messages are cheap right now.🟡 Ending soon. The cheap moment to reply.
Red bar: cache expired, with what a handoff savesOrange bar: cache expired, but the rebuild is cheap
🔴 Expired. What a rebuild costs, and what a handoff saves.🟠 Expired. But the rebuild cost is low.

Features

⏱️ Live countdownMinutes left on the cache, ticking every second; it survives restarts, sleep and switching chats.
💲 Next-message costIn dollars (API prices) or as "× more" (better for subscriptions), plus how full the chat is (290K / 1M).
🚦 One clear tip🟢 keep going · 🟡 reply soon · 🟠 expired small chat, fine · 🔴 expired, with what a handoff would save.
🔔 One pop-upA single warning when a big chat's cache expires, and a heads-up if you send into it. Your message is never stopped.
🔄 /cmaxx-handoffHaiku (effort: High) writes a briefing of the chat; a fresh chat picks up from it. The expired chat is never resent to the main model.
📊 /cmaxx-statsMoney and tokens saved, cache hit rate, rebuilds paid: for this chat and for all chats.
⬆️ /cmaxx-updateUpdates the plugin, or at least its model and price list, so new models are priced on day one.
🔒 Local by defaultNo network and no model calls, except the two commands above, and only when you run them.

Install

Requires a recent Claude Code: the desktop app, the VS Code extension, or the CLI. It's tested on 2.1.325 and later.

Any application
Just paste this repository's link (https://github.com/5m0k3/cache-maxxing) in the chat window and ask Claude to install it for you.

or see below for your specific application:

Desktop app

  1. Click + next to the chat box, then Plugins → Manage Plugins.
  2. On top right press + Add → Add marketplace → Add from a repository and enter 5m0k3/cache-maxxing.
  3. Add plugin → cache-maxxing.

Terminal or VS Code (at the Claude Code prompt)

/plugin install cache-maxxing --marketplace 5m0k3/cache-maxxing

Answer y to add the marketplace and press Enter for the user scope. The bar shows after your next message (may need a restart).

Usage

CommandWhat it does
/cmaxxHide or show the bar in this chat (the ✕ on the bar hides it too).
/cmaxx-handoffWrite a briefing of this chat with Haiku and start a fresh chat from it. (Doesn't trigger the old chat's model)
/cmaxx-statsShow savings, cache hit rate, rebuilds and replies: this chat and all chats.
/cmaxx-updateUpdate to the latest version, or refresh the model and price list.

What each light means, the exact texts and the rules behind them: How the bar decides.

Settings

In the plugin's settings, under cache-maxxing:

SettingDefaultWhat it changes
Rebuild limit ($)0.50Above this rebuild cost, the bar suggests a handoff instead of carrying on.
Warning time (minutes)15How early the bar turns 🟡 before the 1-hour cache expires.
Long chat size (tokens)300000When the 🟢 bar starts suggesting a handoff once you're done.
Show cost asdollarstimes shows "40× more" instead of dollars, which suits subscription plans.
Memory timeautoauto is 1 hour (every Claude subscription chat). Choose 5 with an API key.

Keep it updated

Plugins from GitHub don't update by themselves: Claude Code auto-updates only Anthropic's own marketplaces by default.

  • Once: in a terminal session, /plugin → Marketplaces → cache-maxxing → Enable auto-update.
  • Any time: run /cmaxx-update. If a newer version is out, it updates the plugin (in the terminal) or tells you how (desktop app, VS Code). Either way it refreshes the model and price list, so a newly released model is priced at once.
  • A chat on a newly launched model ends with new model? try /cmaxx-update.

How the bar decides

From top to bottom, the first rule that fits:

  1. No reply yet in this chat: ⚪️ "cache-maxxing will start after you send a message".
  2. Cache expired or reset, and the rebuild costs more than the rebuild limit: 🔴, suggesting /cmaxx-handoff with what it would save.
  3. Cache expired or reset, rebuild below the limit: 🟠 small chat, fine to keep going.
  4. Under 15 minutes left on the 1-hour cache: 🟡 reply soon.
  5. Active, but the chat is past the long-chat size: 🟢 with a handoff suggestion for when you're done.
  6. Otherwise: 🟢 keep going.
🟢 47:12 min left · next message ≈ $0.01 · 50K / 1M · ✅ Keep going
🟡 12:30 min left · next message ≈ $0.05, later ≈ $1.82 (40× more) · 228K / 1M · ⏳ Reply soon or try /cmaxx-handoff
🔴 cache expired · next message ≈ $2.32 (40× more) · 290K / 1M · 🔄 Try /cmaxx-handoff to save 232K tokens | $2.21

Costs are estimates at standard API list prices, and a floor: a reply that uses tools sends several requests. On a subscription you don't pay these dollars; read them as how much of your usage a message takes. "Active" means very likely active: in our measurements about 2 in 100 messages missed the cache even within the time limit.

Privacy

  • The bar sends nothing anywhere and never calls a model. Everything it shows is worked out on your computer.
  • Two commands reach outside, only when you run them:
    • /cmaxx-handoff sends Claude Haiku an abbreviated copy of the chat (the last 120 messages, long ones cut, tool output left out). The briefing is saved in your project under .claude/handoffs/, which gets its own .gitignore, so briefings are never committed.
    • /cmaxx-update downloads this plugin's version number and price list from this repository and sends nothing. GitHub sees your IP address, as for any download.
  • Stored on your computer: a small record per chat (time of the last reply, chat size, model) for 7 days, and counts for /cmaxx-stats (numbers only, each chat's for 90 days, the totals for good). Never the chat's text!

Limitations

  • If your organization blocks Haiku, /cmaxx-handoff says no handoff was written; the chat is untouched.
  • The handoff's saving is an estimate: Haiku's share and a fresh chat's start are taken from measured medians, priced as Claude Haiku 5.5.
  • Switching model and back within the hour shows "cache reset", although the first model's cache may still be warm.
  • /cmaxx-stats reads an older chat's history from its transcript only up to 4 MiB; a bigger chat that started before you installed the plugin is counted from when the plugin first saw it.
  • A brand-new chat in the desktop app shows the bar only after your first message: the app starts the chat, and its plugins, at that point.

FAQ

Does it cost anything to run? The bar, no. /cmaxx-handoff makes one small Haiku call, typically about a cent at API prices.

I'm on a subscription. Are the dollar amounts real? They're API list prices, a way to compare. Set Show cost as to times to see "40× more" instead.

Contributing

Issues and pull requests are welcome. See CONTRIBUTING.md for setup, tests and the project rules, and SECURITY.md to report a vulnerability privately. Changes are listed in CHANGELOG.md.

License

MIT © 5m0k3

// faq

What is cache-maxxing?

A bar above your chat to see when Claude's prompt cache expires, and what your next message will cost, before you send it.. It is open-source on GitHub.

Is cache-maxxing free to use?

cache-maxxing is open-source under the MIT license, so it is free to use.

What category does cache-maxxing belong to?

cache-maxxing is listed under plugins in the Claudeers registry of Claude-compatible tools.

7 views
★ 14 stars
unclaimed
updated 3 days ago

// embed badge

cache-maxxing on Claudeers
[![Claudeers](https://claudeers.com/api/badge/cache-maxxing.svg)](https://claudeers.com/cache-maxxing)

// retro hit counter

cache-maxxing hit counter
[![Hits](https://claudeers.com/api/counter/cache-maxxing.svg)](https://claudeers.com/cache-maxxing)

// reviews

// guestbook

0/500

// related in Claude Plugins

🔓

A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.

// pluginsmultica-ai/★ 230,631[ claude ]
🔓

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explainin…

// pluginsanthropics/⟨Python⟩★ 156,472[ claude ]
🔓

"CLI-Anything: Making ALL Software Agent-Native" -- CLI-Hub: https://clianything.cc/

// pluginsHKUDS/⟨Python⟩★ 58,467◷ Apache-2.0[ claude ]
🔓

financial-services — a Claude ecosystem project on GitHub.

// pluginsanthropics/⟨Python⟩★ 41,409◷ Apache-2.0[ claude ]
→ see how cache-maxxing connects across the ecosystem