The email MCP server: what your AI assistant can do with email

You open Claude or Cursor, type “design a Christmas promo with a festive hero image and our CHEER20 code, then send it to my VIP customers,” and it happens — the template gets built, the email goes out, and your assistant reports who opened it. No dashboard, no copy-paste, no code.

That’s what an email MCP server does. Here’s how it works, what it’s good for, and how to run it.

What an email MCP server is

The Model Context Protocol (MCP) is an open standard that lets AI assistants call real tools. An email MCP server exposes email as a set of tools your assistant can use — so instead of you opening a dashboard, writing the copy, and clicking send, you just ask, and the assistant does it.

The magic is the protocol layer: one server, every client. Set it up once and it works in Claude Desktop, Claude Code, Cursor, Cline, Zed — and any client that ships tomorrow. No custom integration per assistant, no SDK to wire up, no API plumbing.

Why email specifically? Most real work ends in telling a human something — a customer needs a receipt, a user needs a reset link, a list needs a campaign. Email is the assistant’s output channel. Without it, the work stays trapped in a chat window.

What you actually do with it

• Design and send a campaign from a brief. “Design a holiday promo with a festive hero image and our CHEER20 code, then send it to my VIP customers.” The assistant builds the template, sends it, and reports opens — no designer, no dashboard.
• Tailored lifecycle email. The assistant looks at how someone signed up and sends a welcome that actually fits, instead of a one-size-fits-all blast.
• Support replies that write themselves. Read the ticket, look up the order (via another MCP server), draft and send the reply — or escalate when it’s unsure.
• Ops alerts in plain English. Summarise a noisy alert and email the on-call rotation the why, not the stack trace.
• Transactional sends from code. Receipts, OTPs, magic links — fired straight from Claude Code or Cursor while you build.

The pattern: the assistant is the brain, the MCP server is the muscle, email is the output.

Works with the assistant you already use

Claude Desktop · Claude Code · Cursor · Cline · Zed · Continue · and any MCP-compatible client. Set it up once; it works everywhere.

Hosted or self-hosted?

For almost everyone, the hosted server (mcp.brixus365.com) is the answer: point your client at the URL, sign in with OAuth, done. No install, no API key, no infrastructure. (In Claude Desktop: Settings → Connectors → Add custom connector → paste the URL.)

Self-host from source when you want to audit the code, pin a version, run inside an air-gapped or VPC network, or point at a staging environment.

Self-host in five minutes

You’ll need Node 18+ and a Brixus365 API key — the Free plan is 9,000 emails a month, no card.

git clone https://github.com/Brixus-Technologies/brixus365-mcp.git
cd brixus365-mcp
npm install && npm run build

Then point your client at the build (Claude Code shown):

claude mcp add brixus365 -- env BRIXUS365_API_KEY=bx_live_your_key node /path/to/brixus365-mcp/dist/index.js

Ask your assistant “What Brixus365 tools do you have?” — you should see all 16. The full walkthrough is in the self-hosting guide.

Keep it safe

• Start on the Free plan and watch the assistant operate before going live.
• Scope keys per role — emails:send for a sender, templates:write for a designer, marketing:* for campaigns.
• Tag agent-sent mail with a recognisable from_name so it’s easy to spot.
• Keep a human in the loop for large or sensitive sends.

FAQ

Ready to try it? See it on the MCP page, or jump straight to the setup guide and the 16 tools. 9,000 emails a month free, no credit card.