vomit README

The project's README from github.com/zachahn/vomit. As one markdown file: README.md.

Vomit

Vomit converts Claude's token vomit into English by piping it through a local LLM. It's fully local (no telemetry) and has no external dependencies.

Vomit hooks into Claude Code, rewriting all its output within the Claude Code terminal UI.

Disclaimer:

Install

# Install the binary to your GOPATH
go install github.com/zachahn/vomit@latest

# Setup connection details to your LLM
vomit init

# Instructions on how to replace Claude's output via hooks
vomit scrub -claude

Usage

In addition, there's a non-invasive mode if you want to run Vomit on the side. (This is deprecated.)

This should work with:

If you don't have a local LLM set up already, I think I recommend using Llama.app, and downloading GPT-OSS 20B through it. Run the init subcommand after you set this up.

Tips

License

GNU GPLv3

vomit (zachahn/vomit)

Searching Database…

200 ·GitHub
What is this?Vomit converts Claude Code's token vomit into English by piping it through a local LLM. It hooks into Claude Code and rewrites its output right inside the Claude Code terminal UI.
Is this a standalone app or does it hook into Claude Code's terminal?It is a CLI tool written in Go that hooks into Claude Code, intercepting and rewriting its output inside the existing Claude Code terminal UI rather than opening a separate interface.
How do I get it running with Claude Code?Install Vomit, then use commands like 'vomit list' to see Claude session identifiers and 'vomit tail' to translate or follow a session's output. There's also a non-invasive mode to run Vomit on the side, though that mode is deprecated.
What local LLM do I need to run this?You need a local LLM backend such as Llama.app, Ollama, Apfel (macOS 26+), or anything exposing an OpenAI-compatible API. The README recommends Llama.app with the GPT-OSS 20B model, then running the init subcommand once that's set up.
Does the translated output always make sense?The local LLM only sees what Claude tries to communicate, with no access to Claude's actual actions or files, so it can hallucinate. The project is also described as 'totally vibe-coded', only tested on Mac, and noted to be pretty slow.
Does this only work on Mac?It has only been tested on macOS, and one of the suggested local LLM runners (Apfel) specifically requires macOS 26 or later.
Where does my Claude session data go?Vomit runs fully locally with no telemetry and no external dependencies, so session output stays on your machine and is only sent to the local LLM you set up.
How do I get Claude's unmodified output back?Press ctrl-o inside Claude Code to see Claude's original, untranslated output.
What is the non-invasive mode?It lets you run Vomit on the side instead of hooking directly into Claude Code's terminal, but this mode is marked as deprecated in the README.
What does 'vomit tail' actually do?It translates Claude's tokens for a specified session, or follows the latest active session if none is given.
Why does the author call it 'vomit'?The name plays on the idea of cleaning up Claude's verbose, jargon-heavy ('vomit') output into plain English using a local LLM.
Creator
zachahn
Size
0.1 MB
Stars
200
Updated
CommitSep 09
Age
2mo
Commits · 90 Days