Privacy Policy

What we access

To work, reply2mail connects to your mailbox over IMAP/SMTP using credentials you provide, or via OAuth where supported. We read incoming messages to analyze and (in your chosen mode) reply to them.

We store: your account email, agent settings (persona, rules, signature), knowledge you add (RAG sources), calendar events the agent creates, and usage logs.

If you create an API key or connect an AI product, we also store, for each connection: its name, the permissions you chose, its expiry, when and from which IP address it was last used, a log of the calls it made (see "Connected AI products"), and the approval requests it created.

How we protect it

Mailbox passwords are encrypted at rest (AES/Fernet); a leaked database row never reveals your password. Connections use TLS. We never display your password back to you.

We do not sell your data. Email content is processed only to provide the service (analysis, drafting, replies, and passing it to AI products you connect). reply2mail does not use your email content, including content read by a connected AI product, to train or improve any AI model.

AI processing

Message content is sent to our LLM provider (OpenRouter) to generate analyses and replies. Only the content needed for the current task is sent. Sensitive topics (money, contracts, legal) are escalated to you rather than auto-answered.

Requests from an AI product you connect do not go through our LLM provider: reply2mail returns the requested data to that product without generating text with an AI model.

Connected AI products (MCP and API)

You can connect an outside AI product to your reply2mail account: Claude (claude.ai, the Claude apps, Claude Code) or ChatGPT by signing in through their connector, or another tool such as Cursor with an API key you create in the dashboard. Nothing is connected unless you do this yourself.

What flows to the product. While connected, the product can request, and reply2mail returns to it, the data covered by the permissions you granted. With the default permissions this is: your emails in every mail account connected to reply2mail (sender, recipients, subject, date, full body and the text of attachments), your folder list, the email addresses of your connected accounts, your calendar events and your trackers. It does not receive your mailbox passwords, your server settings or your reply2mail login. You can see and narrow these permissions for each connection in the dashboard under Integrations.

Once the product has read your data, that product’s own terms and privacy policy apply to it, not this policy. reply2mail cannot see or control what the product stores, shares, sends through its own tools, or uses for training. Check the product’s privacy settings before you connect it. You direct this transfer; reply2mail makes it on your instruction.

Password-reset and login-code emails. For a connection made from ChatGPT, reply2mail leaves out emails it classifies as login codes, password resets or security codes, and the result says that they were left out. API keys and Claude connections are not filtered: a product with read access can read password-reset links and one-time login codes that other services send you, and a product with web access could use them. The classification can miss an email, so even the ChatGPT filter is not a guarantee. If you would not give a product your mailbox password, do not connect it.

Emails can contain instructions. An email written by someone else can try to tell the connected AI what to do. reply2mail marks email content as untrusted for the AI, but this does not guarantee the AI ignores it. By default, actions that send, forward, reply, archive, move, label or delete mail are not carried out until you approve them in the reply2mail dashboard. Marking emails read or unread and flagging them happen without approval.

Running without approval ("direct mode"). This is optional and available on paid plans only. You switch it on yourself in the reply2mail dashboard, for one API key or connected product and one permission group at a time; to let it send mail without asking you confirm with your two-factor code, or your account password if you do not use two-factor. Its actions in that group then run without an approval request, within limits: with the default settings a send to an address the sending account has not written to before, or above the send limit, still waits for your approval; a send held by the outbound safety filter always waits; and irreversible actions (such as deleting a folder, cancelling an event or removing rules or memory items) always ask. Everything that ran without approval is listed in History on the Approvals page, and you get a notification afterwards that does not contain the email. You can put everything back to approval with one switch. reply2mail’s own assistant and automatic replies are not affected by this setting.

What we log. For every call a connected product makes, we record: the time, which connection made it, the tool used, the outcome and any error code, how long it took, how many items it returned, a reference to the single email it opened or changed (account, folder and message number, not its content), the IP address and a request ID. We do not log the content of your emails, subjects, recipients or what the product asked for. This call log is kept for 90 days and then deleted. You can see recent activity for each connection in the dashboard.

Approval requests. When a connected product asks to send or change mail, we store the request and a preview of it (recipient, subject, text, affected emails) so you can review it. The content of a request is erased 30 days after it is approved, rejected, cancelled or expires; a record that the request existed and its outcome is kept. Security events such as creating a key, changing permissions or connecting an app are kept in our audit log.

Notifications about approval requests do not contain email content.

Webhooks. If you add a webhook endpoint for an API key, reply2mail sends HTTPS requests to the address you give when new mail arrives or an approval request is decided. These requests carry IDs and minimal metadata only (for example an account ID and a message ID), never the subject, sender or body. We store the endpoint address, its signing secret and a delivery log (time, status code, latency, attempt) that you can see in the dashboard.

Keys and connections, and how to revoke them

An API key is shown to you once when you create it. We store only a one-way hash of it, so we cannot show it again or recover it. Keys expire after the period you choose (90 days by default, 180 at most). A connection made by signing in from Claude or ChatGPT stays valid while it is used at least once every 30 days.

You can pause, revoke or disconnect any key or connected product at any time in the dashboard under Integrations. It stops working on its next request, and any approval requests it left waiting are cancelled. Removing reply2mail inside Claude or ChatGPT stops that product from using the connection, but to be sure it is ended on our side as well, also revoke it in reply2mail.

Treat an API key like a password. Anyone who has it can use it within its permissions until it expires or you revoke it.

Your choices

You control the agent mode (off / suggest / auto). You can delete knowledge sources, disconnect your mailbox, or delete your account at any time, which removes your stored settings and credentials.

You decide whether any AI product is connected, which permissions it has, and you can revoke it at any time (see "Keys and connections").

Google user data & Limited Use

When you sign in with Google, reply2mail requests only the basic sign-in scopes (openid, email, profile). We use the result only to identify you and sign you in to your reply2mail account; we keep your email address and do not store Google access tokens. reply2mail does not request access to your Gmail messages through Google OAuth.

A Gmail mailbox is connected with an app password over IMAP/SMTP, the same way as any other mail provider, not through a Google API. Mail read that way is covered by the rest of this policy.

reply2mail’s use and transfer of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements. Specifically: we use Google user data only to sign you in and provide your account; we do not transfer or sell it to third parties for advertising, market research, or any unrelated purpose; we do not use it to train generalized AI/ML models; and no humans read your Google user data except where you explicitly grant permission for support, where required for security or to comply with applicable law, or on data that is aggregated and anonymized.

You can remove reply2mail’s access at any time from your Google Account permissions page.

Contact

Questions about privacy? Email [email protected].