Skip to content
HUIAHUB

rtk-ai/rtk

rtk

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

Metrics updated Established+1.0% 7d
View on GitHub (opens in a new tab)Homepage / demo

GitHub metrics

Stars
83K
+838 this week
Forks
5.3K
30-day growth
—
Open issues
854
82K83K83KOct 4, 2026Oct 13, 2026
GitHub stars from the official GitHub API, recorded each time the repository was checked (3 days).
GitHub stars by day
DayGitHub stars
2026-10-0482315
2026-10-1082829
2026-10-1383153

What it does

RTK (Rust Token Killer) is a command-line proxy that trims the terminal output of common developer commands before an AI coding agent reads it, so less of the agent's context fills up with noise.

It ships as one Rust binary. When an agent runs something like git status, a hook installed by rtk init swaps it for rtk git status, which returns a compact version: failing tests only, grouped lint errors, one-line git confirmations, shortened directory trees. The README lists filters for git, the GitHub CLI, test runners such as pytest, Jest and cargo test, linters, package managers, Docker, kubectl, AWS and Pulumi, and claims more than 100 supported commands with under 10 ms of added time. When a command fails, RTK keeps the full unfiltered output so the agent can fetch it with rtk recall instead of running the command again.

The project is careful about its own numbers. It says it cuts up to 90% of the bash output an agent reads, and states plainly that this does not mean a 90% smaller bill, since shell output is only part of the input tokens. Token counts in rtk gain are estimated as bytes divided by four because RTK has no tokenizer.

Why it matters

An agent reads whatever a shell command prints, including progress bars, passing tests and full diffs. RTK filters that at the source, per command, and the hook means the agent does not need to remember to call it. It covers 18 coding tools according to the README, including Claude Code, Cursor, Codex, Gemini CLI and Copilot. The authors also document where the savings stop, such as built-in agent tools that never pass through the shell.

Main features

  • Compact output filters for git, gh, test runners, linters, Docker, kubectl and AWS
  • Hooks rewrite agent shell commands to rtk equivalents automatically
  • Saves full output of failed commands for later recall
  • `rtk gain` reports estimated savings, with daily, history and JSON views
  • `rtk discover` looks for commands that ran without RTK filtering
  • Native Windows hook since v0.37.2; WSL also works
  • Per-command exclusions and settings in a TOML config file
  • Telemetry is off unless you opt in

Who it's for

  • Developers running Claude Code, Cursor, Codex, Gemini CLI or Copilot on large repositories
  • Teams watching token spend from agent sessions
  • People who want shorter test, lint and git output in agent transcripts

Strengths

  • Works through agent hooks, so commands are filtered without the agent changing its habits
  • Supports a long list of agents, each with its own install flag
  • Honest about what the percentages measure and how token counts are estimated
  • Opt-in telemetry with a documented list of collected fields

Limitations

  • The hook only sees Bash tool calls; Claude Code's built-in Read, Grep and Glob bypass it
  • Savings figures in the README measure shell output, not total cost
  • Token numbers are approximations based on bytes divided by four
  • A different crate named rtk exists on crates.io, so `cargo install rtk` can fetch the wrong tool

Getting started

Homebrew is the README's recommended route, winget covers Windows, and there is a shell script for Linux and macOS. After installing, run rtk init -g (or the flag for your agent) and restart the agent.

brew install rtk
winget install rtk-ai.rtk
curl -fsSL https://raw.githubusercontent.com/rtk-ai/rtk/refs/heads/master/install.sh | sh
cargo install --git https://github.com/rtk-ai/rtk

Requirements: A single binary for macOS, Linux or Windows; some filters call ripgrep (`rg`), which should be on your PATH.

Always follow the official instructions in the README on GitHub. This overview is our summary, not a copy.

Sources