Introducing Mailgun’s open-source MCP server

Accessing email data just got easier. With Mailgun’s open-source MCP server, developers can query email performance metrics using natural language—no dashboards, no complex API calls. Built on Model Context Protocol (MCP), this solution makes AI-powered email analytics seamless and efficient.

Here’s how it happened, and how it works.

MCP, or Model Context Protocol, is a standardized API framework that allows AI assistants to interact with external services using JSON-RPC.

A few months ago, Anthropic released its Model Context Protocol (MCP), enabling AI assistants like Claude to interact with external services through standardized API calls.

In practical terms, it enables natural language interactions with structured data sources like Mailgun’s email analytics APIs.

As developers, we’re constantly looking for more efficient ways to access and analyze our data. Whether checking delivery rates, investigating bounces, or monitoring domain performance, these tasks traditionally require navigating dashboards or writing custom API scripts. MCP opens the door to something different: getting your data by simply asking for it.

Email is unique in being a truly distributed self-governing system. It has been built on open standards that are documented inside the RFCs (Requests for Comments) that are published by the Internet Engineering Task Force (IETF). From SMTP to MIME to DMARC these standards help emails reach their destinations. The ecosystem works because we all agreed on how things should talk to each other.

MCP is built on top of JSON-RPC over an application layer transport, which are well known and documented standards. Adopting it feels like a natural next step following the tradition of open protocols.

By implementing this standard for Mailgun, we’re continuing our commitment to an environment where engineers can build systems without getting trapped in walled gardens. Just as you can send an email to anyone on the planet without thinking twice (please don’t send spam), now you can talk to your data through whatever assistant works best for you.

With our MCP server implementation, you can:

• Query email performance metrics through natural language
• Pull delivery statistics, bounce logs, and domain settings
• Generate reports and visualizations of your email data
• Send emails or validate email addresses

Here’s a glimpse of the kind of metrics you can pull using our MCP and generative AI:

We’ve intentionally limited this implementation to read-only operations (except sending an email). This gives you the freedom to explore your data and configuration without risking accidental changes to your production environment. However, the way it’s designed, it’s easy to enable endpoints that perform non-idempotent requests if your use case calls for it.

We believe infrastructure tools should adapt to how developers get the work done. Conversational interfaces reduce time spent translating questions into complex API queries by using natural language as a bridge.

Mailgun’s MCP implementation reduces cognitive overhead, enabling engineers to:

• Get immediate insights without navigating dashboards
• Request specific email metrics without writing custom code
• Improve workflow efficiency with AI-driven queries

Our MCP server implementation is available on GitHub at github.com/mailgun/mailgun-mcp-server. We’ve included setup steps into the README to help you get up and running quickly.

We’re excited to see what’s next, and welcome contributions and feedback from the community.

Author: Anton Efimenko Anton Efimenko is the VP of Engineering at Sinch Mailgun. He directs the Mailgun Devs and implements cool new email features and service efficiencies. He writes about system updates, services... and one really popular blog post about toasters.