Skip to main content

Homebrew inspector

See what your Homebrew actually does.

Analyse your Homebrew setup: find unused casks, detect conflicts, trace what is on your PATH and which cask you actually open.

I built brewscope because Homebrew knows what you installed, but it doesn't know what you actually use or what conflicts with what. When you move to a new Mac, you want to carry across what matters, not reinstall nine casks that have no app any more.

No installation needed for the interactive demo. Install the CLI with uv tool install.

Abstract amber glass and liquid light
Example output
brewscope
$brewscope describe

Key features

Zero-impact introspection

100%
read-only
JSON
structured output
Zero
side effects
Local
snapshots

What it does.

Homebrew knows what you installed. It does not know that python3 on your PATH comes from pyenv, that two of your casks are both clipboard managers and you have opened one of them twice this year, or that nine casks have no app any more.

brewscope reads the filesystem, your PATH and your usage history to answer those questions.

read-onlyJSON outputAI agent first

Design principles

Built to be safe and transparent.

  1. Read-only

    Never installs, removes or upgrades anything. Recommendations are printed as commands for you to run.

  2. No side effects

    Runs brew with auto-update off, never triggers consent dialogs or bumps access times.

  3. Counts, not content

    Usage is stored as counts per binary. Command lines are never stored or printed.

  4. Local snapshots

    Snapshots stay local because they name your host, paths and dotfiles.

Get started.

Interactive demoTry brewscope in your browser without installing anything.

Install CLI

uv tool install git+https://github.com/rNLKJA/brewscope

About this project

A terminal tool for Homebrew on macOS.

Built for AI coding agents first (Claude Code, Codex, Cursor): JSON on stdout, stable schemas, meaningful exit codes, no prompts. It stays pleasant for humans too, since a terminal gets a table.

Currently at milestone M0: output contract, describe, doctor. Future milestones will add snapshot creation, inventory, conflict detection, usage analysis and migration planning.

What I learned

This project taught me how to design for both humans and AI agents at once. The envelope pattern with stable schemas and meaningful exit codes makes agent integration predictable, while format autodetection (table on a TTY, JSON when piped) keeps it usable for interactive work. I also learned that read-only constraints force better design: when you can't fix problems automatically, your diagnostics and recommendations have to be clear and actionable.

Provenance. This web app is a 2026 revival of a terminal tool built for Homebrew on macOS. Original implementation: github.com/rNLKJA/brewscope. Built with Next.js.