Trust

Security

AgentSentry is a security tool, so we hold our own platform to the same bar we help you measure.

Last updated: June 23, 2026

Stateless by design.Scans run on your machine with your credentials. We never receive your cloud keys and never store your scan results, the data that would be most valuable to an attacker simply isn’t on our servers.

How we protect your data

  • Data in transit, all traffic to the hosted service is encrypted with TLS.
  • Data at rest, we store the minimum (account email, license tier, billing reference). No credentials, no scan contents.
  • Authentication, passwordless magic-link sign-in. There is no password to phish or brute-force. Tokens are single-use, cryptographically random, and expire after 24 hours.
  • Abuse controls, login/signup requests are rate-limited per IP, and responses are generic to prevent account enumeration.
  • Open source, the CLI is AGPL-3.0 and fully auditable. You never have to take our word for what it does.

What the CLI sends home

The only network calls the CLI makes to us are license activation and (for Pro automation) a subscription-status check. These transmit your license key and plan, never your credentials or findings. Cloud API calls go directly from your machine to your cloud provider.

Reporting a vulnerability

If you discover a security issue, please report it privately, do not open a public GitHub issue.

Please include a description and impact, steps to reproduce or a proof-of-concept, the affected component, and any relevant logs or samples.

Scope

In scope

  • agentsentry.org and its subdomains
  • The AgentSentry API (/api/* routes)
  • The nhi-audit CLI package on PyPI
  • This project’s source repository

Out of scope

  • Third-party services we depend on (Vercel, Stripe, Resend), report those to the relevant provider
  • Denial-of-service, spam, or social-engineering attacks
  • Issues requiring physical access to a user’s device

Responsible disclosure

Please give us a reasonable window to investigate and fix an issue before any public disclosure, and do not access, modify, or exfiltrate other users’ data while researching. We’re happy to credit researchers who report valid issues, with your permission, once a fix has shipped.

Machine-readable contact details are published at /.well-known/security.txt.