Vomit is listed as Open Source. Check https://github.com/zachahn/vomit for the current plan limits, as vendors change pricing frequently.

Vomit

Q: What is Vomit?

Vomit is a code AI tool listed on Use AI Tools. A small Go utility that cleans up Claude's token vomit with a separate LLM. It routes verbose model output through a cheaper secondary model so the readable version lands in your context instead of the raw stream, which saves tokens on the expensive model.

Q: What is Vomit best for?

Vomit is commonly used for Heavy Claude Code users watching their token budget, Long agent runs with repetitive narration, Developers already paying for a cheap secondary model. It supports English.

Code🔥 Intermediate✅ Global access

Heavy Claude Code users watching their token budgetLong agent runs with repetitive narrationDevelopers already paying for a cheap secondary model

Open SourceEnglish

A small Go utility that cleans up Claude's token vomit with a separate LLM. It routes verbose model output through a cheaper secondary model so the readable version lands in your context instead of the raw stream, which saves tokens on the expensive model.

Visit Website Compare

Community Rating

Be the first to rate Vomit. Your feedback helps other users decide.

🎯 Real-World Use Cases

Post-process agent output

Send verbose responses to a cheap model for summarising before they re-enter the main conversation.

Cut main-model token spend

Keep the expensive model's context clean so fewer tokens are re-read on every turn.

Readable logs after a long run

Turn a multi-thousand-line transcript into something a human can actually review.

⚖️ Honest Review

✅ Pros

+Small single-purpose Go binary
+Directly targets token cost, the main complaint with agentic coding
+Open source and self-hosted

⚠️ Cons

−Adds a second model call and its own latency
−Depends on a cheap model being good enough to summarise faithfully

📚

Guides and comparisons about code tools

Engineer reviewing terminal output from a coding agent

CodeOct 19, 2026