miross
ESRequest an invite
ProductIntegrationsPricingBlogLog in

Security

How Miross handles your data.

The short version. The long version is the SECURITY.md we keep with the code.

Tenancy

  • Every record carries its workspace; nothing is readable across workspaces.
  • Roles: admin, agent, viewer. Non-members get “not found”, not “forbidden”.

Credentials and keys

  • Passwords are hashed with Argon2id; sessions are opaque tokens stored hashed.
  • Marketplace API keys are encrypted with AES-256-GCM under rotating keys and never returned.
  • Mailboxes and chats connect through a hosted sign-in wizard — credentials never reach Miross.

AI

  • Off by default per workspace. Only stripped message text is sent, only when needed, and every call is metered.
  • Model output is checked before it can reach a customer. Suggested actions are re-verified server-side and confirmed by a person.

Webhooks and email

  • Inbound webhooks (Stripe, Resend, Unipile) are verified by signature or shared secret before parsing.
  • Amazon relay addresses carry a random suffix; only mail from Amazon buyer aliases becomes a case.

Auditing and data rights

  • Append-only audit log of sign-ins, sends, actions and rule runs.
  • Workspace export and deletion are first-class; deletion purges.

Security reports: hola@impulsaia.es