Features

Headroom features

Headroom compresses repetitive input before forwarding requests to Claude Code or ChatGPT Codex. The desktop app manages the local runtime, agent connections, optional add-ons, and usage reporting.

Download for free

7-day free trial · no credit card required

How it works

Local input compression

Headroom compresses repetitive tool output and context locally before forwarding requests to your coding agent's AI provider. Stored originals remain available through retrieval tools. Savings depend on the content processed.

Your tools

logsHTMLJSONshell

Headroom

Compressed input

Claude Code & ChatGPT Codex

receive compressed context

101.8M

tokens saved per developer, on average, after a week of use

Beyond compression

Add-ons, project learnings, and activity

Use optional tools to reduce command output and document reads, save reusable project instructions, and inspect recorded optimizations.

Add-ons

One-click add-ons

Install and enable optional tools from Add-ons. Choose tools for the work you do; core compression works without them.

RTK: summarizes supported terminal-command output.
MarkItDown: converts PDF and Office documents to Markdown.
Serena: provides symbol lookup and targeted code reads.
Codebase Memory: indexes repository structure and relationships.
Context7: retrieves version-specific library documentation.
Ponytail: adds instructions for smaller, simpler implementations.
Caveman: adds instructions for shorter replies.

See Add-ons for setup steps, prerequisites, and tradeoffs.

Headroom add-ons screen: Ponytail, Serena, and Codebase Memory installed with one click

Project learnings

Save project instructions

Headroom scans coding sessions for reusable commands, environment details, and project preferences. It writes saved learnings to the instruction files your agent reads, such as CLAUDE.local.md for Claude Code and ~/.codex/AGENTS.md for Codex.

Use Optimize to review learnings, scan existing session history, and control auto-learning. See the project learnings guide for instructions.

Headroom project learnings: token-saving patterns written to each agent's memory per project

Activity overview

Review recorded activity

Activity lists recorded compressions, learnings, updates, and usage warnings. The dashboard summarizes savings over time. Dollar values represent API-equivalent costs; on a subscription they are not refunds. See how savings are measured.

Headroom activity feed: daily savings records, large compressions, and learnings written to memory

All local, all private

Get started with Headroom

Compression runs locally before requests are forwarded to your AI provider. Follow the quickstart to install and verify routing, then use the tutorials to configure a workflow. See the privacy policy for data handling details.