Hi Product Hunt! I'm Abdulrahman, founder of Hashling.

AI agents like Claude Code, Codex and Cursor can now run programs, read files and push code on our Macs. I wanted a simple answer to "what is each one allowed to do?", and that grew into a personal security agent that explains everything in plain words.

What makes it different:

🤖 It audits the AI agents on your Mac: their permissions, MCP servers, keys left in config files, and live network use.

💬 Your own AI becomes your security engineer. Connect Claude through MCP and ask "is my Mac safe?", or switch on Fully local and a free model runs on your Mac, so nothing leaves it.

🧨 It catches booby-trapped repos, the cloned projects whose settings make an AI agent run code the moment you trust the folder.

🔐 It can only look and suggest. Every change waits for your Touch ID, and most can be undone.

It also tracks what goes in and out of your Mac: how much each app uploads and to where, first contact with new destinations, where every download came from and who signed it, plus API keys and secrets left in plain text. And it checks texts and links for scams, scans for malware, watches startup items, permissions and trackers, and checks your phone over USB.

Being upfront: it isn't a firewall, and it scans files shortly after they arrive rather than blocking them first. The compare page says where other tools do better: https://gethashling.com/compare

It's free and MIT-licensed, with no account and no tracking. Download the installer, double-click it, and you're set up in a few minutes. I'd love your feedback, especially on what's confusing.