gglucass/headroom-desktop — reverse-engineered prompt
Reverse engineered prompt
Build me a local first menu bar desktop app called Headroom for macOS 14 on Apple Silicon, with an experimental Linux build for just the core proxy flow. The whole point is to help people get a lot more Claude Code and Codex usage for the same money by routing those clients through a local token saving pipeline, without changing their normal workflow.
When the app launches, it should set up and manage its own isolated Python runtime in app storage, connect Claude Code and Codex to a local proxy, and make careful backups before editing any user config. I want a clean dashboard that shows tokens used, estimated dollars saved, daily and monthly trends, per client stats, and whether the pipeline is healthy. Include optional local add ons for compressing noisy command output and cleaning up documents before they reach the model.
It should be very transparent about what it changes, easy to pause, and on quit it should remove the routing so Claude Code and Codex act normal again. Uninstall should fully clean up. Look up current docs online if you need to.
Want more depth? Deep Reverse