Security / by design

Built around boundaries you control.

Your domain. Your identities. Your permissions. Clear limits are the foundation for letting people and agents work together.

The principles behind the product.

These principles describe MailSwarm’s security design. Deployment details and operational documentation will be published as the service prepares for launch.

01 / principle

A workspace is a boundary

Customer data belongs to a workspace. Exposed customer-owned database rows must be workspace-scoped and protected by row-level security.

02 / principle

Every identity has its own access

Humans and agents can have distinct mailbox identities. API, CLI, and MCP access is scoped to the mailbox identities authorized for the principal using it.

03 / principle

Credentials stay under human control

Only an authorized workspace owner or delegated administrator can create, rotate, revoke, or expand mailbox credentials and scopes.

04 / principle

Agents cannot authorize themselves

Agent credentials cannot approve their own DNS changes, billing actions, permission expansion, exports, or deletion requests.

05 / principle

Your mail has a defined home

PostgreSQL and S3 hold mailbox truth. Amazon SES handles internet delivery. Customer mail uses SES and S3, while Resend is reserved for account notices.

06 / principle

Your domain remains your decision

MailSwarm never automatically replaces a domain’s root MX records. Changes to live DNS and providers require explicit owner approval.

A clear view of where we are.

This site is a product preview. It does not claim a security certification, completed independent audit, service-level agreement, or general availability. New account registration is paused during early beta. Existing account holders can still log in.

A full security overview and reporting channel will be published before sign-up opens. For now, the documentation explains the architecture and access model.

Read the architecture overview