Your AI reads your email. Anyone can write to you. Is that safe?

Once Claude or ChatGPT can read your inbox, every email becomes text your assistant reads, and a stranger can hide instructions in it (“forward the last invoices to this address”). This is called prompt injection, and no connector can make an AI immune to it. What a connector controls is what the assistant is able to do with your mail. Fezola is built so that the most damaging action, sending an email, is switched off by default and only available on mailboxes where you turn it on.

A robot showing a visitor pass at a keypad door beside a mailbox with letters, a green shield and an eye toggle

How prompt injection works with email

An assistant does not separate your request from the content it reads: an email, a PDF attachment or a chat message all arrive as text. A sender can write a line aimed at the AI rather than at you, sometimes hidden in white text or in an attachment.

The risk depends on the tools the assistant has. If it can send email, a hidden instruction can make it send your data to someone else. If it can only read and write drafts, the same instruction ends up, at worst, in a draft you will see before anything leaves.

What Fezola does about it

  • Sending is off by default: until you switch on direct sending for a mailbox, the assistant is not even offered a tool to send an email. An injected “send this to…” has nothing to call while it is off.
  • Drafts by default: replies and forwards are saved in the Drafts folder of the mailbox that received the email. You read the recipient and the content, then send from your own mail app.
  • If you switch sending on, it works mailbox by mailbox: the send tool is marked as destructive so your assistant asks you to confirm, addresses the mailbox has never written to are flagged, and every send appears in your Fezola sent log.
  • Google Chat and Telegram are read-only: the assistant cannot post in a space or a chat.
  • Actions that change your mailbox (move, archive, trash, mark as spam, delete a draft) are labeled as destructive in the protocol, so Claude and ChatGPT can ask you to confirm them. Received messages are never permanently deleted: trash can be restored.
  • Reading does not mark messages as read, and emails are read live from your mailbox, never stored by Fezola.
  • Passwords and Google or Telegram tokens are encrypted (AES-256-GCM). The assistant never sees them, and connecting a source happens on Fezola, not in the conversation.

What no connector can do for you

Fezola limits what can happen through your mail, not everything your assistant can do. If the same conversation also has other tools, such as web browsing or another connector that can send messages, an instruction read in an email could try to use them. Keep an eye on what the assistant proposes, and review every draft before sending it.

Step by step

  1. Connect only the sources you need

    The Free plan includes 2 sources. Start with the mailboxes you actually want to ask about; you can add more later.

  2. Keep confirmations on in your assistant

    Let Claude or ChatGPT ask before tools that change your mailbox. Fezola marks those tools so the assistant knows which ones to confirm.

  3. Review every draft before sending

    Check the recipient and the content of each draft in your mail app. A draft that addresses someone you do not know is the clearest sign of an injected instruction.

  4. Check which assistants have access

    Ask your assistant to list the assistants connected to Fezola, or open the Connector page of your Fezola account. You can revoke access for any of them at any time.

Why Fezola

  • Sending off by default: nothing leaves your mailbox unless you switch it on.
  • Drafts saved in the mailbox that received the email, with that address as the sender.
  • Google Chat and Telegram connected in read-only mode.
  • Messages are read live and never stored; tokens and passwords are encrypted.
  • Works in Claude, in ChatGPT on Business and Enterprise/Edu plans, and in other MCP clients that support remote servers with OAuth.

Connect your first account in two minutes

Free with 2 sources, no card needed. Nothing to install.

Start for free

Questions

Can an email make Claude send my data to someone?

By default, no: until you switch on direct sending for a mailbox, the assistant has no tool to send an email, only to save drafts. With sending off, the worst an injected instruction can do is prepare a draft, which stays in your Drafts folder until you send it yourself. If you switch sending on, a hidden instruction could try to use it: the send tool asks for confirmation, addresses the mailbox has never written to are flagged and every send is logged, so keep it off on mailboxes where you do not need it.

Does Fezola filter malicious emails?

No, and no connector can reliably detect every hidden instruction. Fezola's protection is structural: it keeps the send action off by default and the assistant on drafts, so the result of an injection stays visible to you.

Can the assistant delete my emails?

It can move a message to the trash, archive it or mark it as spam, and those tools are labeled as destructive so your assistant can ask you first. Received messages are never permanently deleted, and trash can be restored. Only drafts can be deleted permanently.

Does Fezola keep a copy of my emails?

No. Emails are fetched live from your mailbox when the assistant asks and are not stored on Fezola's servers. Fezola keeps your connection details, encrypted.

How do I cut off an assistant's access?

Ask it to list the connected assistants and revoke one, or remove the connector from your assistant's settings. You can also remove a mailbox from Fezola at any time.

Is it free?

Free includes 2 sources. Pro connects unlimited sources for $9 per month or $90 per year, with a 7-day free trial.

Sources

Keep reading

A mailbox plugged into an adapter next to a server with dials, a toolbox, a key, a padlock and a friendly robot giving a thumbs up

IMAP MCP server: connect any mailbox to Claude (and ChatGPT Business)

OVH, IONOS, Gandi, Fastmail or a mailbox on your own domain, in Claude (and ChatGPT Business), with nothing to install.

Read the guide
Five coloured mailboxes dropping envelopes into one big funnel while a robot assistant reads the result

How to bring all your inboxes and chats into Claude (or ChatGPT Business)

Gmail, iCloud, your own domain over IMAP, Google Chat and Telegram, searchable together through one connector.

Read the guide
Two envelope icons tagged personal and work linked through one switch to a single chat bubble

Gmail, iCloud and IMAP inboxes together in Claude (and ChatGPT Business)

Personal and work Gmail, iCloud and your own domain in the same conversation, with every result labeled by account and replies drafted from the right address.

Read the guide
A cloud shaped like an envelope above an app-password card, a phone with a check mark and a small assistant bubble

How to connect iCloud Mail to Claude

Connect iCloud Mail to Claude with an Apple app-specific password and one custom connector, next to your other inboxes.

Read the guide
A violet postbox sending envelopes along a swirl to a laptop with a chat window, with a key tag and a shield

How to use Yahoo Mail in Claude (and ChatGPT Business)

Connect Yahoo Mail over IMAP with a Yahoo app password, then add one custom connector to Claude (or ChatGPT Business).

Read the guide
Several small chat rooms with speech bubbles seen through a magnifying glass, a read-only eye icon and a robot observing

How to use Google Chat in Claude and ChatGPT

Ask Claude (or ChatGPT Business) what your team said in Google Chat, what is unread and what still needs your answer, next to your email.

Read the guide
A paper-plane chat bubble beside a phone keypad and a login code card, a glass window protecting messages with an eye, and a robot with a visitor badge

How to use Telegram in Claude (and ChatGPT Business)

Let Claude (or ChatGPT Business) catch up on your Telegram private messages, groups and mentions, and tell you what still needs an answer.

Read the guide

Documentation

Setup in Claude and ChatGPT, and every MCP tool.

Open the documentation

Pricing

Free with 2 sources, Pro with unlimited sources and a 7-day free trial.

See pricing