Overview

The SenderKit MCP server lets an AI assistant — Claude, Cursor, Windsurf, or any other MCP-compatible client — operate SenderKit through tool calls. Once connected, an agent can send a test notification, scaffold and edit templates, and inspect message delivery without ever leaving your editor or chat.

​

Who it’s for

• Developers using AI coding tools. Wire SenderKit into your assistant once; after that, “send the welcome email to user@example.com” is one line in the chat.
• Agent authors. Building an autonomous workflow that needs to send notifications? Point it at SenderKit over MCP and let it dispatch through the same templates and key prefixes your app already uses.

​

What you can do with it

The server exposes a small surface that mirrors SenderKit’s main operations:

• Send a templated message, or raw inline content, to one recipient.
• List and inspect templates by slug.
• Query and cancel messages — list recent sends with filters, fetch a single message, or cancel a still-pending one.
• Check your connection mode with senderkit_context — confirm whether the active connection is live or test before sending.

All in test mode by default — use an sk_test_ key and your assistant can experiment without anything leaving the building.

​

How it runs

Three ways to connect: The hosted and local options are API-key authenticated and behave identically — pick whichever fits your setup. OAuth is for clients that handle auth themselves rather than accepting a custom header.