Beam Sentinel

Review risky prompts and commands before they are sent to major AI assistants.

Beam Sentinel checks what you're about to send to an AI assistant — before it leaves your browser. Pasting a config, a stack trace, or a customer record into an AI chat takes one click. Beam Sentinel adds a half-second pause before that click when what you're sending looks risky: a live API key, a cloud credential, a private key, an email address, or a command that uploads or exfiltrates data. Ordinary prompts go through with zero friction. How it works - Runs automatically on the major AI chat assistants listed under this extension's site access. - Checks the message right before you hit Send, Submit, Run, or press Enter. - High and critical findings (live-looking secrets, private keys) pause the send — you choose "Keep blocked" or "Allow once." - Medium findings (emails, phone numbers, other labeled personal data) show a short warning and continue. - Everything else sends normally, with no interruption. Local by default - All checks and redaction run inside the extension, on your machine. - There is no account, no sign-in, and no network request — Beam Sentinel doesn't send your prompts anywhere, including to Beam. - Only the last 100 redacted decisions are kept in local browser storage so you can review what was caught. Raw prompt content is never stored. Why it matters AI assistants are part of daily work now, which means secrets and customer data increasingly get pasted into them by accident. Beam Sentinel is a free, lightweight guardrail for exactly that moment, from the team behind [Agentbeam](https://agentbeam.com). Permissions Beam Sentinel only requests the `storage` permission, used to remember your on/off setting and the local action log. It does not request access to your browsing history, other tabs, or any network permission.