privacycommand

drop an app. see everything it touches.

Forensic permission audits for any macOS app.

Warning

Pre-1.0 — no stable release yet. Anything can change in any release, including a patch: APIs, CLI flags, config keys, file formats, and data already on disk. Keep your own backups. Project status. The badge above is generated from the privacykey status list, which says what I promise for this project and every other one.

---

privacycommand takes a .app bundle (or a .dmg) and reports what the app actually touches: the entitlements it claims, the permissions it will ask for, the domains and URLs compiled into its binary, the third-party SDKs it ships, the login items and helpers it registers — and, if you let it, what it does while it runs.

It is built for people who want evidence rather than a vendor's word: security teams, IT, and anyone deciding whether a download deserves a place in /Applications. Every finding carries a plain-English explanation from an in-app knowledge base, so a report is readable without a reverse-engineering background.

All analysis happens on your machine, and the app ships no analytics of its own. The exact list of network calls it makes is in docs/PRIVACY.md.

What it does

• Static analysis — entitlements, code signing (the 10-character Team ID expanded to the developer's name), a notarization deep-dive (stapler / spctl / SHA-256), URL schemes, document types, hard-coded domains, embedded launch agents and helpers, feature-flag and trial-state strings, secrets and licence-key names, anti-analysis signals, dylib hijacking surface, and Apple's Privacy Manifest checked against what the binary actually uses.
• SDK fingerprints — which analytics, advertising and attribution SDKs the bundle ships, with a heat-graded count and per-category breakdown.
• Outbound call sites — which functions can open a connection, the networking symbols they reach for (BSD sockets, getaddrinfo, CFNetwork, nw_*), and any host or URL literals sitting next to them. Any call site can be decompiled on demand if you have Ghidra installed.
• App Store privacy labels — for Mac App Store bundles, the developer's declared Privacy Nutrition Labels sit next to the static findings, so you can see whether the claims match the binary.
• Background Task Management — every login item, launch agent, daemon and helper the app has registered, read through the privileged helper.
• Monitored runs — launch the inspected app under privacycommand and watch, in real time, its file events (via the optional helper running fs_usage), network destinations with reverse-DNS labels, child processes, pasteboard / camera / microphone / screen-recording activity, USB device interactions and resource usage.
• Network kill switch — cut the app off from the destinations it is contacting, via a pf anchor installed by the helper, and watch how it copes.
• VM mode — a guest agent that runs inside a macOS VM (VirtualBuddy / UTM / Parallels) and ships observations back to the host, for apps you would rather not run on bare metal.
• Compare runs — diff any two saved reports from the History tab. Added and removed entitlements, domains, SDKs, login items and findings are colour-cued, with a "show only changes" toggle.
• Batch scan — point it at a folder, or at all of /Applications, and triage many apps at once in a sortable table of risk tiers, warning counts and headline signals.
• Reports — every finding exports as JSON, HTML or PDF.

Get it

Requires macOS 14 or later.

Homebrew — the cask lives in privacykey/homebrew-tap:

brew install --cask privacykey/tap/privacycommand

brew upgrade --cask privacycommand keeps it current. When privacycommand detects it is running from a Homebrew Caskroom it disables in-app updates, so brew stays in charge of the on-disk version.

Direct download — take the signed and notarized .dmg from the latest release and drag the app to /Applications. In-app updates use Sparkle 2 against an EdDSA-signed appcast feed; automatic checks are off by default and you opt in under Settings → Updates.

Command line — the app also carries a command-line tool of the same name, privacycommand, which runs the same analyser from Terminal. There are two ways to put it on your PATH:

• Homebrew — the cask links privacycommand for you on install, along with its zsh, bash and fish tab completion, and removes them on uninstall. Nothing else to do.
• Direct download — choose privacycommand ▸ Install Command Line Tool…. It links the copy inside the app into /usr/local/bin, so the tool updates with the app. When that folder needs administrator rights, as it does on most Macs, it shows the sudo ln -s command to paste into Terminal instead, with a button that copies it. The same menu item then reads Uninstall Command Line Tool… and removes the link.

privacycommand slack --short   # one-line verdict for an installed app
privacycommand preview         # check outdated Homebrew casks before you upgrade
privacycommand upgrade --max-risk medium   # apply the low-risk upgrades, review the rest
privacycommand --help          # every command and flag

Tab completion — Homebrew sets it up. Otherwise add one line to your shell's startup file; privacycommand completion --help lists them. For zsh, put this in ~/.zshrc after compinit:

eval "$(privacycommand completion zsh)"

preview inspects Homebrew casks before you update them. It runs read-only brew queries but never brew upgrade, never blocks an update, and exits 0 whenever it completes. upgrade --max-risk <limit> turns that into a gate: it downloads each incoming build, and a cask whose build scores at or below the limit (low, medium, high, critical, or a score 0–100) is upgraded through brew, while anything riskier is held for you to review — on a terminal it asks about each held app, and the exit status says whether something is still held. Many people keep it behind an alias, alias update='privacycommand upgrade --max-risk medium', so a routine update only stops for the apps worth a look. To build the CLI from source instead, see CONTRIBUTING.md.

Docs

There is no docs site yet. What exists lives in the repo:

• .github/ARCHITECTURE.md — the targets, why they are separate, how data moves between them.
• privacycommand/README.md — source-tree map, signing and entitlements reference, troubleshooting.
• privacycommand/HELPER.md — privileged helper bundling, signing and verification.
• privacycommand/docs/GUEST_AGENT.md — VM guest-agent walkthroughs.
• docs/PRIVACY.md — every network call the app makes, and why.
• NOTICES.md — third-party notices.

Contributing

Issues and pull requests are welcome. CI runs two workflows on every pull request: the SPM build and test suite plus a privacycommand CLI smoke test, and an unsigned Xcode build of the app target. Reproduce the first locally from privacycommand/:

swift build
swift test

Or just build and just test from the repo root. Setup, the Xcode path, and what to include in a pull request are in CONTRIBUTING.md.

Found a security issue? Please do not open a public issue — .github/SECURITY.md has the reporting address and what is in scope.

Licence

Released under the MIT licence.