How imado protects your mail

• Every connection to imado is encrypted (TLS). Mail between imado and other mail systems is sent and received over TLS where the other side supports it, which is nearly everywhere.
• Stored mail, meaning the parsed messages, the attachments, and the original raw message, is encrypted at rest by the providers that hold it: Amazon S3 for the raw messages, PlanetScale for the database.
• Passwords are stored hashed, never in a form anyone can read. We cannot tell you yours.

Nobody at Zivaro Software Inc. can sign in as you. The tooling has no impersonation capability, and there is no support screen that opens a mailbox. The operator's view of the product shows counts, what is in flight, domain names, and who holds an account — never a message.

The people who operate the service can reach the underlying storage, because somebody has to be able to repair it. The privacy page says when they do.

• Every notification our mail provider sends about a message (arrival, delivery, bounce, complaint) has its signature verified before anything is recorded from it.
• Our mail provider checks every arriving message for spam and viruses. A new conversation whose first message fails either check starts in Spam; a failing reply stays in the conversation it answers, so a conversation you are in never disappears.
• A message's HTML is shown inside a sandboxed frame on its own origin. Scripts, forms, embedded frames and every event handler are removed before it is rendered, and the frame is allowed to fetch images and nothing else.
• An attachment is served as a download, never as a page. A picture under 10 MB (PNG, JPEG, GIF or WebP) is also shown on its message, drawn as an image and nothing else. The file you download is checked against a fingerprint (a SHA-256 digest) recorded when the message arrived.
• Addresses you type are parsed under a strict grammar before any message is sent.
• Mail to an address that nobody created reaches no mailbox. Our mail provider accepts it and we record it, and the sender gets no bounce. The exceptions are postmaster@ and abuse@, which reach the oldest address on the domain so a person reads them.

• An account needs a verified email address before it can do anything.
• Sign-in attempts are limited to 10 a minute per network address, sign-ups to 10 an hour, and password-reset mail to 3 in 10 minutes. The limits are shared across every server, so they cannot be dodged by hitting a different one.
• Your Account page lists every browser signed in to your account, and signs out any of them, or every other one at once.
• Resetting your password signs out every browser and ends every app you connected.
• A cookie keeps you signed in. Your theme preference stays in your browser. So does a copy of your recent mail, encrypted with a key that is destroyed when you sign out.
• A suspended account is refused at sign-in, with a sentence saying how to reply.

• An app such as Claude or ChatGPT connects through MCP, the standard way AI apps sign in to a service. You choose which mailboxes it reads and whether it may write drafts or manage your setup, and you can disconnect it from your Account page.
• No app can send mail. An app can write a draft, every draft it writes is marked as written by that app for everyone who reads the mailbox, and a person presses Send.
• An app reads the text you would see, as plain text. Text hidden in the usual ways is dropped, though a sender can still hide some from you that the app reads. Attachments are listed by name and size and never opened, and reading never marks anything read.
• Anyone can email you, so a message can be written to steer an app that reads it. imado tells the app that mail is information from its sender, never an instruction from you, but an app can still be fooled. Check who a draft is addressed to before you send it.
• An app that manages your setup cannot remove anything, invite people or spend money, and your Account page lists each change an app made.
• imado checks an app's permission on every request it makes, so disconnecting it, or resetting your password, stops it at its next request.
• In a mailbox you share, everyone who reads it sees which apps can read it and who connected each one.
• Our logs record which tool an app used and how many items it got back, never the text of your mail.
• What an app does with what it reads is up to the app and the company that runs it. If it can also browse the web or use other connectors, text it read in your mail can go there, and imado cannot see or stop that.

The guide to connecting an assistant says how to connect one and everything it can do.

• Two-factor sign-in. Today a verified address, a hashed password and rate-limited sign-in protect the account.
• Scanning attachments ourselves. Today our mail provider's virus check is the only one, and a file is served exactly as it arrived, as a download or, for a picture, as an image, so nothing in it runs in your browser, and opening it is your decision.
• Loading images through our own servers, so a sender sees us rather than you. Today they load from the sender, as in most mail apps — the privacy page says what that means.
• An external audit and a disclosure programme. Until then, this page names what is built, and the address below is read by the people who built it.

When one of these lands, this page changes with it.

Write to ashish.so@zivaro.ai with “security” in the subject. The same address is published at /.well-known/security.txt. You will hear back within two working days, with what we found and when it was fixed.

Please do not read, change or delete mail that is not yours while demonstrating a problem, and do not run anything that degrades the service for other people. A report made that way is welcome, and we will not pursue anyone for making one.