rtkCLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies

rtk

rtk is a high-performance CLI proxy that dramatically reduces LLM token consumption by filtering and compressing command outputs before they reach your AI assistant's context. Built as a single Rust binary with zero dependencies, it intelligently rewrites 100+ common developer commands—from git and file operations to test runners—achieving 60-90% token savings with less than 10ms overhead.

Key Features

• Massive Token Reduction: Cuts token consumption by 60-90% on typical development workflows through smart filtering, grouping, truncation, and deduplication strategies
• Zero-Friction Installation: Single Rust binary with no external dependencies; installs via Homebrew, curl, or pre-built binaries for macOS, Linux, and Windows
• 100+ Supported Commands: Optimized filtering for git, file operations, test runners (pytest, cargo test, jest), linters, GitHub CLI, and Docker with auto-hook integration for Claude Code, Copilot, Cursor, and other AI agents

Use Cases

• AI-Assisted Development: Accelerate Claude Code, GitHub Copilot, and Cursor sessions by reducing noise in command outputs that clutter LLM context windows
• Cost Optimization: Lower API bills and improve response times by minimizing unnecessary tokens sent to LLM providers during code reviews and debugging
• Large Codebase Navigation: Maintain context efficiency when working with large projects where standard commands generate verbose, token-heavy output

Who Is It For

rtk is designed for developers and teams using AI coding assistants who want to maximize context window efficiency and reduce LLM API costs. It's particularly valuable for those working in large codebases or running frequent development commands within AI-powered workflows.