Secrets are encrypted
Mailbox passwords, OAuth tokens and connected-integration keys are encrypted before they are stored, using AES-256-GCM with a random nonce per value. Encryption and signing use separate keys derived from one master key.
Passwords for MailStream accounts are never stored. Only a salted scrypt hash is kept.
Tokens and keys are hashed
Session tokens and API keys are stored as SHA-256 hashes. A copy of the database does not contain a usable session or API key, and an API key is shown in full only once, when you create it.
Unsubscribe links, confirmation links and webhook deliveries are signed, so they cannot be forged or altered.
Tenant isolation
Everything a customer creates (mailboxes, domains, leads, campaigns, suppressions, API keys) belongs to one workspace, and server queries are scoped to the signed-in workspace. Users can belong to several workspaces and act in one at a time, with roles inside each.
The database also enforces row-level security on every table with no public access policies, so its public data API cannot read or write customer data; only the application server reaches it.
A separate, audited admin app
Platform administration lives in its own application with its own sessions, which expire on an absolute timer. Changes made by platform admins are recorded in an audit log.
Rate limits and bot protection
Rate limits are stored in the database, so they hold across several server instances. They cover sign-in attempts (with backoff), the public API, form submissions, mailbox connections, click tracking and the AI endpoints.
Public sign-up forms use a honeypot field to catch simple bots, and Cloudflare Turnstile verification when it is configured.
Outbound connections are guarded
Because MailStream connects to servers you name (SMTP, IMAP and websites to crawl), those connections are checked against private and internal network ranges, and the validated address is pinned for the connection, to prevent server-side request forgery.
Security headers
Every page is served with these response headers:
- Content-Security-Policy with object-src none, a restricted form-action and frame-ancestors none
- X-Frame-Options DENY (hosted forms are the one place embedding is allowed)
- X-Content-Type-Options nosniff
- Referrer-Policy strict-origin-when-cross-origin, and no-referrer on pages whose URLs carry secret tokens
- Permissions-Policy that turns off camera, microphone and geolocation
- Cross-Origin-Opener-Policy same-origin
Hosting region
The application and admin functions run in Frankfurt (EU), pinned next to the database.
What this page is
This is a list of protections that exist in the product today. It is not an audit report or a certification. If you need a specific answer for a security review, email support from the contact page.