The most common request we got after launching AI email generation was some version of: "Can I just tell it to send?" Not generate a draft, not give me a subject line to copy — actually send. So we built the MCP integration. Sendflo now works as an MCP server, which means you can connect it to Claude Desktop, Cursor, Windsurf, Zed, or any MCP-compatible AI and run your email marketing from there.

This is not a chatbot that writes copy for you to paste somewhere else. When you tell your AI to send a campaign, it sends the campaign. Contacts get created. Segments get queried. Emails go out. You stay in one window.

What MCP actually is

MCP (Model Context Protocol) is an open standard for connecting AI tools to external services. The short version: instead of each AI app building its own Sendflo integration, we publish one server and every MCP-compatible app can use it. Claude Desktop, Cursor, Windsurf, Zed, Cline, GitHub Copilot — they all work with the same config.

The difference from previous AI email tools is execution. Earlier tools could draft copy or suggest subject lines. An MCP tool reads your real contact list, creates a real campaign, and hits send. The AI is not generating suggestions. It is operating your account.

What you can do with it

Claude Desktop
You
Send a New Year sale email with 40% off to all VIP contacts
Calling generate_email → "New Year sale 40% off", tone: friendly
Calling list_segments → found "VIP Customers" (847 contacts)
Calling create_email_campaign → "New Year Sale 2026"
Calling send_campaign → campaign_id: 91
Done. Your New Year sale campaign is sending to 847 VIP contacts. Subject: "🎉 Happy New Year — 40% off, today only".

Four tool calls. One message. No dashboard, no segment ID copied, no send button clicked.

Other things you can ask:

The 11 tools available

Eleven tools, each mapped to a real action in your account:

How to set it up

You need a Sendflo account, an API key, and an MCP-compatible app. Claude Desktop, Cursor, Windsurf, and Zed all work. The config is identical across all of them.

Step 1: Install uv

uv handles the Python runtime. You don't need to install Python or manage versions.

curl -LsSf https://astral.sh/uv/install.sh | sh

Step 2: Get your Sendflo API key

Go to app.sendflo.io → Settings → API Keys → Create. Copy the key starting with sk_sf_.

Step 3: Add Sendflo to your AI app config

Open your MCP config and paste this:

{ "mcpServers": { "sendflo": { "command": "uvx", "args": ["sendflo-mcp"], "env": { "SENDFLO_API_KEY": "sk_sf_your_key_here" } } } }

Where to find the config file:

Step 4: Restart and go

Quit your app fully and reopen it. In Claude Desktop, Sendflo tools show under the hammer icon. In Cursor and Windsurf, check the MCP panel in settings.

Full setup guide

The complete setup guide with screenshots is at sendflo.io/docs.

Auth

No new login, no OAuth. The MCP server uses the same sk_sf_ bearer token that powers the Zapier integration and REST API. Create one under Settings → API Keys, drop it in the config, done.

Agency accounts: each workspace gets its own API key and its own config entry. The list_workspaces tool returns all sub-workspace IDs if you need to look one up.

Where this fits in practice

The dashboard isn't going away. Building templates, configuring sending domains, managing billing — that still happens there. What the MCP integration removes is the constant switching for operational stuff: checking stats mid-conversation, adding a contact someone just mentioned, sending a campaign you already planned.

If you already live in Claude, Cursor, or Windsurf most of the day, you've got one fewer context switch. The more interesting use case is automation — triggering campaigns off external events through an AI agent, rather than manually. A deal closing in your CRM, a signup hitting a threshold, a product launch date arriving. The MCP server handles the Sendflo side; how you wire the trigger is up to you.

Try the MCP integration

Set up in 5 minutes. Works with Claude, Cursor, Windsurf, and more.

Read the setup guide →