All your inboxes in Claude Code and Claude Cowork

Short answer: to add all your inboxes to Claude Code, create a free Fezola account, add your mailboxes, then run claude mcp add --transport http fezola https://api.fezola.com/mcp and type /mcp in Claude Code to sign in. Claude can then search Gmail, iCloud and any IMAP mailbox from your terminal, with every result naming its account. The OAuth part works in production; a full end-to-end test with a real account is still pending, so treat this as "should work" and tell us if something breaks.

Why email in a coding agent

Developers already live in the terminal. Having Claude Code check the thread where a client reported a bug, find the invoice for a cloud bill, or draft a reply while you work saves a context switch. The problem is that a built-in connector usually covers one provider. If you have a Gmail, a work mailbox on your own domain and an iCloud address, you want one connector that holds all of them. That is what an email MCP server does.

What you need

• Claude Code installed, with a recent version that supports remote HTTP MCP servers.
• A free Fezola account with at least one mailbox added in Sources (Gmail, iCloud, OVH, IONOS, Gandi, Yahoo or any IMAP mailbox).
• A browser on the same machine, for the OAuth sign-in.

Step 1: add your mailboxes to Fezola

Create the account, open Sources and add each mailbox with a short name ("Personal", "Work"). Your assistant uses that name to say where each message comes from. Providers that need an app password, such as iCloud, are explained in the all your inboxes guide.

Step 2: add Fezola to Claude Code

Per the Claude Code documentation (checked on 8 November 2026), a remote HTTP server is added with:

claude mcp add --transport http fezola https://api.fezola.com/mcp

By default the server is stored for the current project only (local scope). To use it in all your projects, add --scope user:

claude mcp add --transport http --scope user fezola https://api.fezola.com/mcp

Step 3: sign in with OAuth

Inside Claude Code, run:

/mcp

Select Fezola and follow the browser sign-in. You land on the Fezola consent screen, which lists the sources the assistant will access: accept, and the tools become available. The same login can also be started from your shell with claude mcp login fezola.

What works today, checked by Fezola in production: the client registers itself, receives the authorization page and reaches the consent screen. What is still pending: a complete run with a real account, from the first command to the first answer.

Step 4: check and ask

claude mcp list

shows Connected when it works, Needs authentication if you still have to run /mcp. Then try:

• "What came in on all my mailboxes today that needs an answer?"
• "Find the email where the client reports the checkout bug, in any mailbox."
• "Find the invoice PDF from the hosting provider and tell me the total."
• "Draft a reply to the last email from my accountant, from the right address."

Behind the scenes the agent calls tools such as search_messages, get_thread, get_attachment and reply_draft. Attachments in PDF, Word, text and image formats can be read.

Claude Cowork

According to our information, Claude Cowork reaches the connectors you add in Claude itself (Customize, Connectors, + Add, Add custom connector, paste https://api.fezola.com/mcp). A Fezola connector added there should therefore be available in Cowork. We have not tested this end to end, so check your own setup and tell us what you see.

What a coding agent should and should not do with your mail

• It cannot send by default. Fezola saves drafts in the mailbox that received the message, with that address as the sender. The send_message tool exists but is off by default on the server. A coding agent that runs commands should not be able to email people on its own.
• Email is untrusted input. A message can contain instructions aimed at the model. Since the connector cannot send, a hijacked agent cannot mail anyone, but still review what it proposes.
• You can cut the access. In Claude Code: claude mcp logout fezola then claude mcp remove fezola. In Fezola, list_connected_assistants and revoke_assistant show which assistants are connected and revoke them.

Other editors and agents are covered in email in Cursor, VS Code, Windsurf and Codex CLI. For a comparison of connector options, see best email MCP servers, and for safety, is it safe to connect your email to an AI?.