MCP for AI Assistants

The Model Context Protocol (MCP) is an open standard that enables AI assistants to securely connect with external services. Lettr provides two MCP servers so AI agents can work with your email infrastructure in plain language: sending and scheduling messages, managing templates, domains, and webhooks, running campaigns, managing your audience, and monitoring delivery.

​

Why Use MCP?

MCP transforms how you interact with your email infrastructure. Instead of switching between dashboards, writing code, or memorizing API calls, you can ask your AI assistant to handle email tasks directly.

​

Two Ways to Connect

Lettr’s MCP server is hosted at https://app.lettr.com/mcp and accepts either a Lettr sign-in or a lttr_ API key. A second server, the lettr-mcp npm package, ran on your own machine — it is no longer updated.

​

Comparison

​

Which Should I Use?

Use the remote server in most cases. It works in web chats like Claude.ai and ChatGPT as well as in desktop apps and coding tools like Claude Desktop, Claude Code, and Cursor. You sign in with your Lettr account, so there are no API keys to create or store, and it has the widest set of tools. Where a browser sign-in isn’t possible, such as CI, cron jobs, or headless agents, connect to the remote server with an API key instead. The local server is deprecated. It still works, and its code is still open, but it will not gain anything further. If you are running it, switching takes one command and the same API key.

​

What Can You Do?

Both MCP servers give AI assistants tools to work with your email infrastructure. Each tool does what the matching API endpoint does. See the tools reference for the full list.

​

Example Conversations

Once connected, you can interact with your email infrastructure using natural language: Sending emails:

“Send an email to john@example.com from hello@mycompany.com about tomorrow’s 3pm meeting”

Using templates:

“Send the ‘order-confirmation’ template to buyer@example.com with order_id ‘12345’ and customer_name ‘Sarah’”

Debugging delivery:

“Why didn’t my message to jane@example.com deliver yesterday?”

Checking domains:

“Is my domain notifications.example.com verified and ready to send?”

Reviewing events:

“Show me all bounces from the last week”

Scheduling:

“Schedule the ‘weekly-digest’ template to team@example.com for Monday at 9am”

Managing your audience:

“Add these five addresses to the ‘Beta testers’ list and subscribe them to the Product updates topic”

Running campaigns:

“How did last week’s newsletter campaign perform?”

​

Security

Both MCP servers are designed with security in mind:

• Remote server — Authenticates through OAuth 2.1, or with a lttr_ API key limited to that key’s permissions. Every tool runs as you, in the team you choose (see Choosing a Team), and goes through the same team permission checks as the dashboard. API keys are never exposed to the AI assistant.
• Destructive tools — Tools that send email, overwrite data, or delete it are marked as destructive, so your AI client can ask you before using them. The tools reference marks each one.
• Local server — Authenticates with an API key stored in your local environment. The key never leaves your machine. Use restricted API keys to limit what the AI can do.

​

Next Steps