02 / Go further · one line at a time
~/.config/rtk/pipeline.toml
on by default
Less command output
Test runs, diffs, logs and builds are compressed before your agent reads them. Every agent.
up to 90%less command output
$ rtk init -g
one line
Every tool, not just the shell
Grep results, web pages, lockfiles and any output rtk has no filter for get the same treatment.
Claude Code
extends = "rtk:posthook"
bring a model
Big files, read for you
A read over 350 lines is answered by a cheaper model you choose. The file never enters your agent's context.
Claude Code · your API key
extends = "rtk:delegate-reads"
optional daemon
Clean API traffic
Noise stripped from the conversation your agent re-sends every turn, with the prompt cache kept intact.
any agent you point at it
$ rtk claude
Install a pipeline someone shared, or share yours
$ rtk pipeline add github:user/repo · $ rtk pipeline share publish
−20.2%total cost · author's benchmark
best listed on rtk-ai.app/pipelines
best listed on rtk-ai.app/pipelines