Connecting your account with Claude (Cowork): automatic tasks and live dashboards

Picture opening your AI assistant in the morning and typing: "how are my campaigns doing this week? Build me a dashboard and let me know if any of them opened below average." And having it actually do it — with your real data, instantly, without you touching the platform or writing a single line of code.

That's what you unlock by connecting arrobaMail to an assistant like Claude (in its Cowork, Desktop, or Code mode) through MCP, an open standard that lets AI read and operate your account within whatever permissions you define. In this guide we connect it end to end and show you what to ask it. It's the most "futuristic" tutorial in the set, but the connection itself is surprisingly simple.

1. What you'll be able to do

Today, checking your campaigns means opening arrobaMail, clicking around, staring at numbers, and putting together a summary by hand. With this integration, you talk to the AI and it does all of that for you: it queries your data in real time, cross-references it, builds you a dashboard, flags what matters, and even drafts campaigns for you to review.

It doesn't replace Amanda AI — the intelligence that lives inside arrobaMail — it complements it. Amanda produces and analyzes inside the product; this integration lets you operate your account from your own assistant, alongside the rest of your work.

2. How it works: the players in the flow

Before connecting, it's worth understanding who does what. There are four players in a chain, and data flows between them live:

• You ask for something in plain language.
• Claude (Cowork) interprets the request and figures out what data it needs.
• MCP is the bridge: it translates that request into something arrobaMail understands, authorizes it with your API Key, and enforces the permissions.
• arrobaMail returns your up-to-the-moment data: campaigns, lists, metrics, reputation.

What matters most: the AI reads your data and prepares things (like drafts), but it never fires off sends. That boundary is built in by default, and we cover it in step 7.

3. Get your API Key

The API Key is what authorizes your assistant to act on behalf of your account. You generate it from the dashboard, under Tools › API Key, and copy it.

Important: treat your API Key like a password. Don't paste it anywhere public and don't share it. If you ever suspect it leaked, you can regenerate it from that same section, and the old one stops working instantly.

4. Connect arrobaMail to your assistant

With the key in hand, connect arrobaMail to your assistant. The process varies a bit depending on the client, but the logic is always the same: you give it arrobaMail's MCP server address and your API Key. In Claude, for example, it's a single command; in other clients, a small configuration step.

We won't reproduce the exact commands here because they can change depending on your client and your account: you'll always find them up to date on the MCP page and in your own dashboard, under the MCP (AI) tab, with specific instructions for Claude, Cursor, VS Code, and Windsurf. Follow those steps, and once you're done, your assistant will confirm arrobaMail is available as a tool.

Tip: if you use Claude daily, this connection takes less than two minutes and saves you from opening the platform for every question. Connect it once and it stays connected.

5. Ask for your first live dashboard

This is where it gets fun. With the connection made, ask your assistant something concrete. For example:

"Show me the performance of my campaigns from the last 7 days in a dashboard: sent, opens, clicks, and the trend by campaign."

Your assistant queries arrobaMail live, organizes the data, and builds the dashboard for you. And because it's connected, it's never a stale snapshot — every time you ask, it pulls your data fresh. That's what makes it "live." You can ask it to narrow the focus ("just the campaigns for my customer list"), compare it to the previous week, or flag anything that's out of the ordinary.

6. Automate recurring tasks and reports

A dashboard is just the beginning. Since your assistant already "sees" your account, you can hand off tasks you used to do by hand:

• Recurring summaries: "every morning, give me a rundown of how my active campaigns are doing."
• Adding contacts: "add these three emails to the beta testers list" — without ever leaving your work tool.
• AI-generated drafts: "prepare a campaign draft announcing the new feature, professional tone, for the active customers list." The AI creates it as a draft in your account; you review and send it from the platform.
• Quick audits: "which of my lists has the most engagement?" and the AI cross-references the data for you.

Each of these — things that used to take several clicks and a lot of navigating — is now just a sentence.

7. Security: you authorize, you stay in control

Giving an AI access to your marketing doesn't mean giving up control. If anything, the guardrails are built in from the start:

• You hold the key. Without your API Key, there's no access. And you can revoke it whenever you want.
• Permissions by type. Every tool is labeled either Read or Write, so you always know exactly what it can query and what it can touch.
• It never sends on its own. The most sensitive action stays in your hands: the AI prepares drafts, but you're the one who hits send.
• Open standard. MCP doesn't lock you into one provider: what you connect today, you can keep connecting tomorrow.

The full detail on the security model and available tools lives on the MCP page.

Common mistakes to avoid

• Sharing or exposing your API Key. It's the key to your account. If you're ever in doubt, regenerate it.
• Expecting the AI to send campaigns. It won't, on purpose: it prepares drafts, you approve them. That's a safeguard, not a limitation.
• Making vague requests. "Show me data" doesn't get you far; "build a dashboard with opens and clicks from the last 7 days, by campaign" gets you a lot. The clearer the ask, the better the result.
• Skipping the dashboard's setup steps. Per-client instructions live in your account and on the MCP page — follow them to get the connection right the first time.

Next steps

1. See every available tool and the security model in MCP.
2. If your business needs custom integrations beyond the assistant, check out connecting arrobaMail to your systems.
3. Pair it with in-platform analysis: understand your reports with Amanda AI.