MCP serversSkillsPlugins & marketplacesHooks

headroomlabs-ai/headroom

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

Stars
80,265
+7 days
—
Forks
6,284
Language
Python
License
Apache-2.0
Last push
22 Nov 2026

Summary

Summary written by AI from the README; it may be imprecise.

This tool compresses context such as tool outputs, logs, files, and conversation history locally before sending it to language models. It runs as an in-process library, a local proxy, an MCP server, or an agent wrapper for tools like Claude Code to reduce token consumption. The system also supports reversible compression via local caching and shapes verbosity to lower output token costs.

Use it when

  • When passing large logs, JSON payloads, or tool outputs to an LLM and needing to cut token usage
  • Wrapping Claude Code or other coding agents to compress context without altering application code
  • When requiring reversible context compression with local caching across multi-agent workflows

Install

pip
pip install headroom-ai
pip
uv tool install --python 3.13 "headroom-ai[all]"

Commands quoted from the repo's README. Read the README before running them.

Includes hooks that run shell commands on your machine. Read them before installing.

Stars over 90 days

Collecting: 2 of 90 days so far. The chart appears after 7 days.

Type to search the reference, templates and lessons.