← All articles

How to Use Claude Code as Your Email Assistant (No API Fees)

2026-08-09

If you already pay for Claude Code, you are quietly holding an email assistant you have never fully used. Most people reach for Claude Code to write code or documents, but the very same assistant can connect to your inbox over the Model Context Protocol (MCP) and take over the most tedious parts of email — reading, searching, and drafting. And it does all of this with no extra API charges: everything runs inside the plan you already pay for.

In this article we'll look at why "no API fees" matters so much in practice, what you can actually ask Claude Code to do with your mail, how to connect it in a few minutes, and how the design answers the obvious worry — "will the AI send something on its own?"

Why "no API fees" changes everything

Most "AI email" tools work like a taxi meter. Behind the scenes they call a model API, and every summary, search, and draft consumes tokens you pay for. Crucially, that bill usually sits on top of the AI subscription you already have, so you pay twice for the same intelligence. Zephmail removes that structure entirely by reusing the Claude Code or Codex you already run.

Email is something you touch dozens of times a day. Even a few cents per action adds up quickly across every message, every teammate, every month. Worse, a visible meter changes behavior: people hesitate to ask the assistant for help because each request costs something. An efficiency tool that makes you flinch every time you use it slowly stops getting used.

MCP flips this around. The mail client never calls an AI API. Instead, it behaves as a local server that your own Claude Code connects to. Your plan does the thinking; the mail app is just the doorway. So no matter how many times you summarize or redraft, there is no additional invoice from the mail app.

What you can actually ask Claude Code to do

Once connected, you simply talk to it in plain language. Here are some of the most common requests:

• "Summarize my unread mail in three lines." — Get the shape of this morning's inbox without opening a single message.
• "Sort by urgency and tell me only what I need to answer today." — Not just a list, but a prioritized triage.
• "Find the latest message from the finance team." — Full-text search finds it even when you only half-remember the subject or date.
• "Draft a polite reply to this invoice thread saying payment will land at month-end." — A reply that quotes the original appears in your composer.
• "Remind me what this person sent me last week." — Instant recall of past correspondence.

The key point is that the assistant is answering by reading your real mail. You never paste messages into the chat. Through MCP tools it pulls genuine data from your live mailbox and bases its summaries and drafts on it — which is why it is far less likely to hallucinate an email that doesn't exist.

Connecting it (done in minutes)

There is no complicated configuration. Roughly, it's three steps:

1. Install Zephmail and add your email account. For Gmail, you just paste an app password (server settings fill in automatically).
2. Start the server from the "AI integration (MCP)" panel. The app shows a connection URL, a token, and a ready-made command for Claude Code.
3. Copy the one-line claude mcp add … command into Claude Code. That's the whole connection.

Once connected, you just talk to Claude Code and the results show up in Zephmail. The connection is local (127.0.0.1) with token authentication, so nothing outside your machine can reach it.

The design that prevents "accidental sends"

When you hand email to an AI, the biggest fear is a message going to the wrong person on its own. Zephmail's answer is unambiguous — the AI drafts, you send.

When the assistant writes a draft, it merely appears in your composer; nothing is sent. You read it, fix what you want, and send it yourself when you are satisfied. Even when the assistant proposes a send, that request only lands in an approval queue and never leaves your outbox until you press approve. Deleting and archiving work the same way — the AI never does them alone. Speed comes from the AI; the final decision stays with the human.

A day in the life

Picture an afternoon between focused work. You open Zephmail and Claude Code is already connected. "Tell me only the unread from this morning that probably needs a reply," you ask, and it narrows things to a handful with the key points. "Draft a first response saying I'll confirm with the responsible team and get back to them," you add, and a courteous message is ready for you to check before sending. "Summarize the invoice thread from last week in three lines," you continue, and a long exchange snaps into focus. Even in fragments of time, your inbox never piles up — email gets handled without lifting your hands off the code you were writing.

Common questions

Is my Claude Code plan really enough? Everyday email actions — summarizing, searching, drafting — are lightweight and comfortably fit within normal coding usage (actual limits depend on your plan). Either way, there is no separate charge from the mail app.

Does it work with multiple accounts? Yes. Zephmail supports multiple accounts and a unified inbox, with per-account signatures and smooth switching, so you can handle work and personal mail together.

Can I use Codex instead of Claude Code? Yes. Codex also supports MCP and connects the same way. Whichever you choose, sending stays approval-gated and there are no extra API fees.

An interface that is just "asking"

The biggest appeal of using Claude Code for email is that you don't have to learn the tool. No filter syntax, no search operators, no keyboard shortcuts. You say what you want in ordinary words. "Where's that invoice from last week?" "Tell this person next week is fine." Claude Code translates those natural requests into the right email operations. If traditional mail software was a tool you had to master, AI integration is a tool you simply ask. The learning curve is close to zero, so you work at full speed from day one.

This ease helps the most for people who struggle with email. If you can never keep folders tidy or you let replies pile up, a single "summarize my unread and tell me what to answer" gives you the whole picture. You don't need a perfect filing system. Leave the inbox messy, let the AI find what matters, take the summary, and have it draft the replies. Instead of disciplining yourself with rigid rules, you clear things by asking as you go — that is what living with email looks like in the AI era.

Synergy with your coding work

The more you already use Claude Code for code and documents, the more valuable email integration becomes, because your development context and your email context now live inside the same assistant. You can make requests that cross both: "Round up these GitHub notification emails and tell me which PRs are waiting on review," or "Find the customer bug reports and pull out just the reproduction steps." There's no tool-switching; you process email inside the same conversation, keeping your working context intact.

This is a direct benefit of MCP being an open standard. Claude Code treats your email tools as one capability among many, side by side with other tools. So requests that combine several sources just work, with no special integration effort.

Getting set up without stumbling

A few things smooth out the first connection. When adding your account, if you're asked for an app password, use the dedicated 16-character code rather than your normal login password (for Gmail, two-step verification is required). When registering with Claude Code, copying the one-line command the app shows is the reliable route — typing it by hand invites token mistakes, so paste it.

If something doesn't work after connecting, it's almost always one of two things: the MCP server isn't running, or the token is stale because the server was restarted. Check that the server is running in the AI integration panel and, if needed, grab the latest command again.

The value of using it without watching the meter

"No API fees" is more than a price point. When every action is billed, people unconsciously ration themselves. "Do I really need this summary?" "Redrafting again might cost more." Those small hesitations pile up until a capable assistant gathers dust. Remove the surcharge and the brake comes off: you summarize freely, redraft until you're happy, and try anything that occurs to you. An efficiency tool only becomes a daily habit when you can use it without counting the cost — and because everything stays inside the Claude Code plan you already pay for, you can lean on it without holding back.

Where it really pays off

Concrete scenes make the value obvious. Morning triage: grasp overnight arrivals in three lines and plan your day. Long replies: hand off the skeleton of a careful refusal or a fact-checked answer that would take ages to write from scratch. Finding things: surface "that quote from last month" by full-text search even with a fuzzy subject. Cross-language work: get a natural English draft for an overseas contact by conveying just the key points. Handovers: summarize the history of a thread to share when responsibilities change. Each of these happens constantly, and even a few minutes saved per instance compounds into real time.

Common misconceptions

"If the AI reads my mail, is the content used for training?" Zephmail does not send your mail to its own cloud. Processing happens on the plan of the assistant you subscribe to, and the connection is local only. How data is handled follows the policy of the assistant you use.

"It sounds hard." The real work is "start the server and paste one line," after which you just talk in plain language. You never write commands or scripts.

"Won't it just add a checking step?" You do review drafts, but that's far faster than writing from scratch, and checking before you send is something everyone does anyway. What's added is one safety click; what's removed is the whole writing effort.

"Won't I get too much mail for even AI to keep up?" The opposite — high volume is where the benefit is largest. Checking every message yourself is a slog, but hand filtering and summarizing to the assistant and a busier inbox is grasped faster, not slower.

Wrapping up

The Claude Code subscription you already pay for becomes a powerful email assistant with a little connection work. With no extra API fees, you can hand off summarizing, searching, and drafting, while sending always stays under your approval. Because coding context and email context now share one assistant, everyday information handling gets even smoother. "I have a brilliant assistant, yet email is still manual" — that's the state MCP-native Zephmail is built to end.