Features
Everything the Headroom app does
Headroom is a local proxy that reversibly compresses the logs, boilerplate, and repetitive context before they reach Claude Code or ChatGPT (previously Codex) (~50% fewer tokens on the JSON and log items it compresses, typically 40-50% across a real-world agentic coding session, with no measurable hit to quality), so the plan you already pay for goes further. The app wraps that engine in tools to see, extend, and act on the savings.
7-day free trial · no credit card required
How it works
Less noise in, more code out
Headroom runs as a local proxy: it intercepts each prompt before it reaches Claude Code or ChatGPT and reversibly compresses the logs, boilerplate, and repetitive context that bloat it, keeping the original retrievable on demand. The JSON and log items it compresses shrink by ~50% with no measurable hit to quality; across a real-world agentic coding session that typically adds up to 40-50% fewer tokens overall, automatically.
Beyond compression
Three features that compound the savings
Compression trims every prompt automatically. These build on top of it, extending the savings to your terminal and documents, teaching your agent to stop repeating expensive mistakes, and showing you exactly what's being saved.
Add-ons
One-click add-ons
A small set of optional tools you install and toggle with a single click, wired up for both Claude Code and ChatGPT with no separate setup. Each attacks a different slice of the token bill, and they stack:
- RTK trims noisy terminal output, so your agent sees the important parts of a command's results without the clutter.
- MarkItDown converts PDF and Office documents to Markdown so they cost far fewer tokens to read.
- Serena gives your agent symbol-level code tools, so it reads the one function it needs instead of the whole file.
- Codebase Memory indexes your codebase into a persistent knowledge graph, so structure questions are answered from the map instead of by re-reading files.
- Context7 fetches current, version-specific documentation for the libraries you use, so your agent stops burning tokens on guessed or outdated APIs.
- Ponytail nudges the agent to write the least code that solves the problem: the output side of the bill that input-focused tools leave untouched.
- Caveman makes the agent reply in terse caveman-speak, cutting output tokens while keeping code, commands, and errors exact.
Project learnings
Turn repeated mistakes into permanent fixes
Headroom scans your past Claude Code and ChatGPT sessions and writes token-saving patterns into each agent's own memory: CLAUDE.md and MEMORY.md for Claude Code, AGENTS.md and instructions.md for Codex. When the agent repeats a mistake, Headroom updates the memory so it doesn't happen again.
Because the patterns live per project, the savings compound: every future session starts already knowing what wasted tokens last time. You can also scan history on demand, or seed learnings yourself.
Activity overview
See every saving as it happens
The activity feed gives you one place to see what Headroom has been doing: compressions and savings recorded, learnings written, RTK saves, available updates, and warnings as you approach a usage limit, so hitting a cap never comes as a surprise. The dashboard rolls it up into total tokens and dollars saved over time, with a daily savings history chart.
All local, all private
Try it on a real project
Everything above runs on your machine; your prompts and code never need to leave it. Download the app, connect Claude Code or ChatGPT, and the add-ons, learnings, and activity feed all light up on their own. The compression engine is the open-source Headroom CLI; see the FAQ for privacy details.
7-day free trial · no credit card required