Vibe coding workflow
Vomit CLI pipes dense Claude 5 output through local LLMs for readable terminal logs
Vomit is an open-source local CLI tool designed to translate dense and verbose token streams from Claude 5 into readable English. By piping Claude's terminal output through a local model like GPT-OSS 20B, it keeps developer sessions uncluttered without external telemetry.
August 21, 2026 4 min read
Curated by Oleksandr Kuzmenko, AI Product EngineerUpdated August 21, 2026Sources cited on every story
AI-assisted · editor-reviewedHow we use AI
Impact: Low
Why it matters
Clean up verbose Claude 5 output streams in terminal coding sessions using zero-telemetry local inference.
TL;DR
- 01Vomit rewrites verbose Claude 5 output streams into concise text using a local LLM.
- 02Runs with zero telemetry and writes temporary execution data strictly to $TMPDIR.
- 03Can run as a terminal hook or sidecar process using vomit tail.
Key facts
- License / Runtime
- Open-source Go binary (zero dependencies)
- Telemetry
- None (100% local processing)
- Target Architecture
- Tested on macOS / macOS Terminal Hooks
Managing Claude 5 Verbosity with Local Pipelines
Vomit addresses terminal output noise in agentic CLI sessions. It intercepts raw Claude token streams and summarizes them locally.
Key Technical Characteristics
- Zero External Telemetry: Runs completely on host hardware with no API calls or remote tracking.
- Zero External Dependencies: Distributed as a standalone Go binary compiled via
go install. - Non-Invasive Operation: Intercepts output at terminal boundaries or runs sidecar via session ID tracking (
vomit tail). - Recommended Engine: Pairs with local runners like
Llama.appexecutingGPT-OSS 20B.
Installation and Usage Commands
# Install Vomit binary
go install github.com/zachahn/vomit@latest
# List active Claude session identifiers
vomit list
# Follow and summarize the latest session stream
vomit tailTry it in 2 minutes
go install github.com/zachahn/vomit@latest
vomit list
vomit tailbash
✓ When to use
- When long Claude 5 reasoning blocks visually flood your terminal prompt during vibe-coding.
- When monitoring live agentic terminal sessions in real time via sidecar logs.
✕ When NOT to use
- When strict, byte-for-byte exact output logs are needed for deep compiler debugging.
- In environments without hardware capacity to run a 20B local LLM alongside agentic IDEs.
What to do today
- Install Vomit using go install github.com/zachahn/vomit@latest to process terminal logs locally.
- Set up Llama.app with GPT-OSS 20B to summarize Claude sessions without cloud latency.
What the community says
“Opus 5 is clearly generating dense tokens for tool calling, but man is it bad for general communication, like borderline unusable.”
“YES. A thousand times yes. It's garbage.”
Sources