rtk-ai/rtk

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

What it solves

RTK (Rust Token Killer) reduces the volume of bash output that AI agents read, cutting up to 90% of the noise. This prevents LLM context windows from being filled with redundant boilerplate, long logs, and verbose command outputs, which lowers costs and improves agent efficiency.

How it works

RTK acts as a high-performance CLI proxy written in Rust. It intercepts shell commands and applies four primary compression strategies:

  1. Smart Filtering: Removes noise like comments and whitespace.
  2. Grouping: Aggregates similar items, such as grouping errors by type or files by directory.
  3. Truncation: Keeps essential context while cutting redundant data.
  4. Deduplication: Collapses repeated log lines into counts.

It can be used as a standalone CLI tool or via an auto-rewrite hook that transparently converts standard commands (e.g., git status) into rtk equivalents (e.g., rtk git status) before the agent executes them.

Who it’s for

Developers using AI coding agents (such as Claude Code, Cursor, Windsurf, or GitHub Copilot) who want to reduce token usage and avoid context clutter when agents execute shell commands.

Highlights

  • Broad Command Support: Supports over 100 commands across files, Git, GitHub CLI, test runners (Jest, Pytest, Cargo), build tools, and cloud providers (AWS, Docker, Kubernetes).
  • Low Overhead: Single Rust binary with less than 10ms of overhead.
  • Transparent Integration: Native hooks for 17 different AI tools to automate command rewriting.
  • Token Analytics: Built-in dashboard (rtk gain) to track estimated token savings over time.

Related

  • Dispatch
    RTK Token Savings: Benchmarks Reveal Limited Cost Reduction in AI Coding

    A comprehensive benchmark of Rust Token Killer (RTK) shows that while it compresses terminal output, it often fails to reduce total AI coding costs due to increased agent turns and misleading 'rtk gain' metrics.

  • Project
    jia-gao/leanctx

    A drop-in prompt compression library that reduces LLM input-token costs by 10-40% using loss-tolerance routing to protect critical code and tool calls while compressing prose.

  • Project
    zdk/lowfat

    A lightweight CLI tool that reduces AI token costs by filtering CLI output and compressing file content before it reaches an AI agent.

  • Project
    edouard-claude/snip

    A CLI proxy that filters verbose shell output before it reaches AI coding assistants, reducing token usage by 60-90% through declarative YAML pipelines.