Headroom compresses tool output before it reaches the model, and keeps it reversible.

headroomlabs-ai/headroom at 62.5K stars sits between an agent and its LLM compressing tool outputs, logs, files, and RAG chunks in flight, claiming 20% fewer tokens for coding agents and 60-95% fewer for JSON at unchanged answer quality (GitHub). Six content-type-specific compressors: SmartCrusher for JSON, an AST-based CodeCompressor covering Python, JS, Go, Rust, Java and C++, a HuggingFace model for prose. The design point that matters is the Compress-Cache-Retrieve store making compression reversible, so the model can request the original span when the summary isn't enough. Four deployment shapes including a zero-code-change proxy on port 8787. Apache 2.0.