CleanMac

A native, private, AI-assisted disk cleaner for macOS.

Scan for caches, logs, duplicates, large files and unused apps — review redacted metadata with an AI CLI you installed — and move them safely to the Trash. CleanMac has no account or telemetry; the AI CLI may use its provider's network service.

English · 简体中文 · 繁體中文 · 日本語 · Español · Français · العربية · हिन्दी · Português (BR) · Русский · বাংলা

---

What is CleanMac?

CleanMac is a native macOS app (SwiftUI) that reclaims disk space safely. It scans your Mac for cleanup candidates — caches, logs, temporary files, duplicates, oversized files, leftover app data — assigns each a risk level, and lets you review everything before anything is touched. Files go to the Trash, never straight to permanent deletion.

What makes it different: it can hand bounded, redacted metadata for selected items to an AI CLI you already installed (Claude Code, Codex, or Gemini CLI) for a second opinion. Your question is sent as written; CleanMac does not add file contents or full paths, and the CLI may contact its configured provider.

Features

• 🧹 Smart scan — caches, logs, temporary files, Trash, developer junk (Xcode derived data), downloads, and unclassified “other” files.
• 📦 Duplicate finder — groups files by SHA-256 content hash and keeps the newest copy.
• 🐘 Large files — surfaces files over a configurable threshold (default 500 MB).
• 🗑️ App uninstaller — moves selected app bundles to the Trash while leaving user support data untouched to prevent accidental data loss.
• 🤖 AI Review (local) — ask an installed AI CLI to classify candidates into safe to delete, risky, and needs review.
• 🛡️ Safety first — 20+ safety rules and three protection tiers block system, Mail/Messages/Safari, and Keychain data from deletion.
• 🌍 11 languages — fully localized UI that follows your system language or a manual override.
• 🔒 Explicit privacy boundary — no CleanMac telemetry or account; AI handoff is redacted and disclosed.

AI Review (Local)

CleanMac detects supported command-line AI tools on your PATH (including common Homebrew, npm, asdf and volta locations) and lets you pick one to review up to 80 cleanup candidates. The app sends your question plus structured anonymous metadata (item ID, size, modified date, category, risk and rule IDs), runs the CLI from a unique empty temporary directory with a 120-second timeout, and accepts the reply only when every item is classified exactly once.

Tool	Binary	Models you can pick
Claude Code	claude	Default · Fable · Opus · Sonnet · Haiku
Codex	codex	Default · gpt-5.5 · gpt-5.4 · gpt-5.4-mini
Gemini CLI	gemini	Default · Pro · Flash

The candidate list and your prompt are passed to the CLI via stdin/arguments as a subprocess — they never leave your machine. CleanMac also strips its own session markers from the child environment so a CLI can't misdetect a nested session.

Privacy & Safety

• AI network disclosure. CleanMac does not implement telemetry, but an installed AI CLI may contact its configured provider. AI review sends your question as written plus anonymous IDs and bounded metadata, never file contents or automatically collected full paths.
• Trash, not rm. Everything is moved via FileManager.trashItem(at:), so you can restore it.
• Protection tiers. allowed (caches/logs/temp) → requiresReview (source code, cloud storage, downloads, dev data) → blocked (system root, app data, Mail/Messages/Safari, browser data, Keychain).
• Full Disk Access is optional but recommended so scans can see protected Library locations. CleanMac guides you through granting it in System Settings.

Install

Download (recommended)

1. Grab the latest CleanMac.dmg from the Releases page.
2. Open the DMG and drag CleanMac into Applications.

Official downloads are signed and notarized. If Gatekeeper cannot verify one, do not disable quarantine; delete it and download the asset again from the official Releases page.

Requirements

• macOS 14.0 (Sonoma) or later
• Apple Silicon or Intel

Build from source

git clone https://github.com/lgqyhm2010/CleanMac.git
cd CleanMac

# Build & run the app bundle (creates dist/CleanMac.app and launches it)
./script/build_and_run.sh

# Or just build with SwiftPM
swift build --product CleanMac

# Run the tests
swift test

Toolchain: Swift 6.0 (Xcode 16+). The package exposes an executable target CleanMac (the app) and a library target CleanMacCore (logic, models, services — kept separate for testability).

Packaging a DMG

A single script builds a universal app and packages a drag-to-install DMG:

# Local, explicitly non-distributable preview
./script/build_dmg.sh --unsigned

# Distributable release; requires Developer ID and notarization credentials
CLEANMAC_VERSION=1.0.0 CLEANMAC_BUILD_NUMBER=1 \
  NOTARY_PROFILE=CleanMacNotary ./script/build_dmg.sh --release

Formal releases fail closed: a missing credential, invalid version, architecture, signing, notarization, stapling, or Gatekeeper check stops the build. Configure release metadata and signing via environment variables; secrets are never hard-coded:

Variable	Purpose
CODESIGN_IDENTITY	Signing identity, e.g. Developer ID Application: Name (TEAMID). Auto-detected if unset.
CLEANMAC_VERSION / CLEANMAC_BUILD_NUMBER	Release version and numeric build number.
NOTARY_PROFILE	A notarytool keychain profile name (see docs/RELEASING.md).
APPLE_ID / APPLE_TEAM_ID / APPLE_APP_PASSWORD	Alternative notarization credentials (used by CI).

--unsigned is only for local/PR validation and is never published as a release. See docs/RELEASING.md for setup.

Pull requests and pushes to main build a read-only unsigned preview. Only a v* tag can publish, after the signed and notarized release passes every check (workflow).

Localization

CleanMac ships in 11 languages: English, Simplified Chinese, Traditional Chinese, Japanese, Spanish, French, Arabic, Hindi, Portuguese (Brazil), Russian, and Bengali. The UI follows your system language by default and can be overridden in Settings.

Strings live in Sources/CleanMacCore/Resources/<locale>.lproj/Localizable.strings behind the L10n abstraction. To add a language, add a new .lproj folder with a translated Localizable.strings, add the case to AppLanguage, and rebuild.

Project structure

Sources/
  CleanMac/          App target — SwiftUI views, stores, AppKit shell, menus
  CleanMacCore/      Library target — models, services, safety rules, localization
Tests/
  CleanMacCoreTests/ Core logic tests
  CleanMacUITests/   App/store tests
script/
  build_and_run.sh   Build the .app bundle and launch it
  build_dmg.sh       Build + package (and sign/notarize) a DMG

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for how to build, test, and open a pull request. Translation fixes and new locales are especially appreciated.

License

CleanMac is released under the MIT License.