agentleFS
ww-w-aiww-w-ai/super-token-saver/llms.txt

super-token-saver

llms.txt35 starsChanged 2 months agoApache-2.0View on GitHub

The only Claude Code plugin built from CC source code analysis. 45% cost reduction measured on real workloads. super-token-saver automatically reduces Claude Code API costs by preventing cache expiry waste, delegating work to cheaper SubTask sessions, and providing real-time cost visibility. Works with both Max Plan ($200/mo subscribers) and API pay-per-use billing. Built by analyzing Claude Code source code (v2.1.88): - Traced token injection to context.ts:36-111 (gitStatus) and tools/BashTool/prompt.ts:53 (commit instructions) - Identified cache structure paths in utils/api.ts:321 (splitSysPromptPrefix) -…

Files
llms.txt
# super-token-saver

> The only Claude Code plugin built from CC source code analysis. 45% cost reduction measured on real workloads.

## What it does

super-token-saver automatically reduces Claude Code API costs by preventing cache expiry waste, delegating work to cheaper SubTask sessions, and providing real-time cost visibility. Works with both Max Plan ($200/mo subscribers) and API pay-per-use billing.

## Key features

- Token Guardian: Detects prompt cache expiry and blocks $9 re-sends before they happen
- Session Architect: Auto-delegates heavy work to SubTasks with 37.5% cheaper cache writes
- Concise Mode: Reduces output token padding while preserving substance
- /s-continue: Replaces /compact with zero-cost, zero-information-loss context restoration (processes 60MB+ in <1s)
- Status Line: Real-time cost/context/rate-limit monitoring at <50ms overhead
- /usage-view: Interactive HTML dashboard with AI-powered cost analysis in 23 languages
- /setup-git-lite: Removes 2,200 hidden tokens CC silently injects every session (~$48/mo saved)

## Measured impact

- $326/day → $180/day (45% reduction) on full-day coding workload
- Cache expiry prevention: $9 saved per catch, ~$270/mo at daily frequency
- Git instruction optimization: ~$48/mo saved
- Plugin overhead: ~1,400 tokens/session ($0.07 per 100-call session)
- Net effect with git-lite: plugin consumes less than it removes

## Technical approach

Built by analyzing Claude Code source code (v2.1.88):
- Traced token injection to context.ts:36-111 (gitStatus) and tools/BashTool/prompt.ts:53 (commit instructions)
- Identified cache structure paths in utils/api.ts:321 (splitSysPromptPrefix)
- Mapped three cache invalidation paths (MCP active, no MCP, 3P provider)

## Installation

```
/plugin marketplace add ww-w-ai/marketplace
/plugin install super-token-saver@ww-w-ai
```

## Links

- Source: https://github.com/ww-w-ai/super-token-saver
- License: Apache-2.0
- Languages: 23 (auto-detected from OS locale)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.