Don't let a leaked API key
destroy your month.

Honestly, you should ask this. We request the minimum possible — read key metadata, create a monitoring rule, and delete one specific key when triggered. No log access, no data access, no ability to create keys ever. If granting us any setup access feels like too much, our Manual Setup option means you create the monitoring rules yourself and only give us permission to delete a specific key. That's it.

Use Manual Setup. You follow our step-by-step gcloud instructions, create the alerting policies yourself inside your own project, and grant CloudSentinel one permission only — delete a specific key when triggered. We never touch your project setup. Same protection, full control on your end.

No — and we built the architecture specifically to prevent this. We use GCP IAM Conditions to scope the delete permission to the exact key resource name. That means even if something went wrong on our end, the permission physically cannot touch any other key in your project.

Budget alerts email you — hours after the damage is done. They rely on billing data which GCP updates slowly, sometimes 4-12 hours behind. By the time the alert lands in your inbox, a leaked key could have already run up thousands. We monitor raw request count instead, which updates in near real-time. When your threshold is crossed, the key is already dead.

It works on any GCP API that uses API keys and shows up in Cloud Monitoring request count metrics — Gemini, Maps, Vision, and most other common ones. A heads up on BigQuery though: its pricing is complex and harder to catch purely through request count. For BigQuery we'd recommend using quota limits alongside CloudSentinel rather than relying on it alone.

You could — and if you want to, you should. This isn't magic. CloudSentinel is for the team that knows the risk is real but doesn't want to build, test, maintain, and debug this themselves. Same reason people use Stripe instead of building their own payments. The core mechanic isn't complex. The value is that it's already running so you don't have to think about it.

The announcement actually validates how real this problem has been. Worth noting though — Google's spending caps only cover the Gemini API. If you're using Maps, Vision, Firebase, Translate, or any other GCP API, there's still no native kill switch. CloudSentinel protects all GCP API keys, not just Gemini. And even for Gemini specifically, spending caps work at the billing level which still lags hours — we monitor raw request count in near real-time and revoke the exact key rather than stopping all your usage.