If you discover a security vulnerability in AgenticMail, please report it responsibly. Do not open a public GitHub issue for security vulnerabilities.

Open a private security advisory on GitHub: https://github.com/agenticmail/agenticmail/security/advisories/new

Or contact the maintainer directly via GitHub.

Include the following in your report:

• Description of the vulnerability
• Steps to reproduce
• Potential impact
• Suggested fix (if any)

1. You will receive an acknowledgment within 48 hours
2. We will investigate and provide an initial assessment
3. We will work on a fix and coordinate disclosure with you
4. Credit will be given to reporters (unless anonymity is preferred)

• Master key — full administrative access to the API
• Agent API keys — scoped per-agent access
• Inbound webhook secret — authenticates Cloudflare Email Worker requests

• Outbound guard — blocks emails containing sensitive data patterns (API keys, credentials, PII) and requires human approval before sending
• Spam filter — scores inbound emails and blocks spam before delivery
• DKIM signing — domain mode emails are signed for authenticity
• SPF/DMARC — DNS records configured automatically in domain mode

• Agent passwords are stored in the local SQLite database
• The database file should have restricted file permissions (0600)
• Cloudflare API tokens are stored in the gateway configuration
• The .agenticmail/ directory contains sensitive configuration and should not be committed to version control

• Stalwart mail server runs locally (Docker)
• API server binds to 127.0.0.1 by default
• Cloudflare Tunnel provides secure ingress without exposing ports
• IMAP/SMTP connections use TLS where available

1. Use strong master keys — generate with openssl rand -hex 32
2. Restrict file permissions on .agenticmail/ and .env files
3. Keep dependencies updated — run npm audit regularly
4. Use domain mode with DKIM for production email sending
5. Configure the outbound guard to prevent AI agents from leaking sensitive data
6. Monitor spam logs for false positives and adjust thresholds as needed