headroomlabs-ai/headroom is a Python-based tool that compresses tool outputs, logs, files, and RAG chunks before they are sent to an LLM, reducing input tokens. The latest release is the v0.41.0 listed in the repository. FACTS contains no specific changelog for this version, so what changed cannot be confirmed. It is offered as a library, a proxy, and an MCP server, and is released under the Apache-2.0 license.
Compression is chosen by format: SmartCrusher handles JSON, CodeCompressor handles code with an AST-based method, and Kompress-v2-base handles prose. CacheAligner flags volatile elements that can break provider key caching, and CCR keeps originals locally so they can be recovered when needed. The README's own benchmark reports that code search dropped from 17199 to 13597 tokens, and SRE incident analysis from 55957 to 24340 tokens. Codebase exploration fell from 58801 to 33895 tokens, and issue classification from 46067 to 32429 tokens. The repository description claims token savings of 20% for coding agents and 60–95% for JSON-family workloads.
These figures are self-benchmark results from the repository description and README, not independently verified performance. The README also notes that the GSM8K score is unchanged at 0.870 before and after compression, while the TruthfulQA score moved from 0.530 to 0.560, but with only 100 samples that cannot be called a meaningful difference. So rather than treating the savings as general performance, it is better to check them on the inputs you actually use.