I use Claude Code all day and got tired of watching it eat tokens on noisy CLI output. git status on a big repo, find across a project, cargo test with 200 lines of compilation — none of that noise is useful to the model.
So I built RTK. It wraps commands and strips the junk before it reaches your context:
rtk git status # instead of git status
rtk find "*.rs" . # compact results
rtk ls . # token-optimized tree
rtk cargo test # just the test results
My rtk gain -q after 15 days of real usage:
Total commands: 7,061
Input tokens: 29.3M
Output tokens: 4.8M
Tokens saved: 24.6M (83.7%)
Biggest surprise was rtk find — directory listings are insanely wasteful. Run rtk discover on your Claude Code session history to see how many tokens you've been wasting — that's what convinced me to actually build this.
Rust, MIT licensed. Happy to hear what commands are the worst token offenders for you.
https://github.com/rtk-ai/rtk