What Muzzle checks
Every check runs on your Mac: inside Claude Code and Codex, on git commits and on the requests your agents' scripts make.
Going out
No keys or emails sent out
Keys, tokens and your personal details stopped in commands, code and requests.
No keys in your commits
A commit that puts a key or your email where the code sends it out doesn't go through.
Find what already leaks
Finds the keys and emails your code already sends out, file and line.
A contact email that forwards to you
A forwarding address for User-Agents, crawlers and push keys.
Coming in
Know when a page gives your agent orders
Instructions aimed at the AI in pages, files and tool results, flagged as they come back.
No malware from installs
Known malware is blocked before your agent installs it. New and unknown packages are flagged.
What other Macs catch, yours flags
Bad hosts and injections caught on other Macs are flagged on yours the next day.
Put Muzzle between your agents and the internet
Free · Apple Silicon · macOS 12 or later
Pro from $8/month, or free while you contribute to the Threat Feed