Get it for macOS

Stop API keys from landing in your agent chats, and find the ones already there. Rejects a message that contains a credential before it reaches the agent or thread history. Scans Claude Code, Codex and BB history on your server in one click. Lists each credential by kind and fingerprint with a link to its revoke page, and never shows, stores or sends a value.


Overview

View source

Pasted an API key into a chat? It now sits in plain text in your agent history. Token Guard stops the next one before it is sent and shows you the ones already there.

What you get

  • Messages containing a credential are rejected before they reach the agent or the thread history.
  • A Token Guard page scans Claude Code, Codex, and BB history in one click.
  • A report with the kind of each credential, a fingerprint, where and how often it appears, and a link to revoke it.

How it works

Everything runs on your BB server. Values are never shown, logged, or stored, and nothing leaves the machine. Only formats with a distinctive prefix are recognized, so false blocks stay rare. It does not check attached files or other enrolled machines.

More from Alexander Kashin

More in Security 5

Protect credentials or prevent unsafe code.

View all