LeanCTX — Reduce Claude Code, Cursor & Copilot Token Costs by 99%

# LeanCTX — Reduce AI Coding Token Costs by 99% > Open-source MCP server + shell hook that reduces Claude Code, Cursor, GitHub Copilot, and Windsurf token costs by up to 99%. Single Rust binary with MCP tools, 90+ CLI compression patterns, and zero telemetry. Works with 24 AI coding tools. ## What is LeanCTX? LeanCTX (lean-ctx) is an open-source context compression system for AI coding tools. It sits transparently between your AI tool (Cursor, Claude Code, GitHub Copilot, Windsurf, Gemini CLI, etc.) and the LLM, compressing all context before it reaches the model. This reduces token usage by 60-99%, cuts API costs, and makes AI responses faster. Install in 60 seconds: `npm install -g lean-ctx-bin && lean-ctx setup` ## How It Reduces Token Costs 1. **MCP Server**: Compression-aware tools replace built-in file reads, shell commands, and code search. Uses tree-sitter AST parsing across supported languages. Read modes + caching; cached re-reads cost ~13 tokens. 2. **Shell Hook**: Intercepts 90+ CLI command patterns (git, npm, cargo, docker, kubectl, pip, go, tsc, curl) and compresses output by 60-95%. 3. **AI Tool Hooks**: One-command integration for 24 editors via `lean-ctx setup`. No workflow changes needed. 4. **Communication Protocols**: CEP (Cognitive Efficiency Protocol), CCP (Context Continuity Protocol), and TDD (Token Dense Dialect) teach the AI to respond more efficiently, saving 8-25% additional tokens. ## Real Token Savings | Operation | Without lean-ctx | With lean-ctx | Saved | |:---|---:|---:|:---:| | File reads (cached) | 30,000 tokens | 195 tokens | 99% | | git status/log/diff | 8,000 tokens | 2,400 tokens | 70% | | ls / find | 6,400 tokens | 1,280 tokens | 80% | | Test runners | 10,000 tokens | 1,000 tokens | 90% | | curl (JSON) | 1,500 tokens | 165 tokens | 89% | | **Typical session** | **~89,800** | **~10,620** | **88%** | Active developers save **$30-100+ per month** on AI API costs. ## Scientific Compression Engine The compression engi