Build an MCP Server in 10 Minutes or Less

MCP (Model Context Protocol) opens up a powerful new way to let AI clients interact with your backend data and logic. In this guide, you'll learn how to set up an MCP server in Xano, create tools that perform specific actions, and connect everything to an AI client like Claude Desktop — all in under ten minutes.

Setting Up Your MCP Server

To get started, navigate to the AI Tools section in the left-hand menu of your Xano workspace and create a new MCP server. You'll give it a name, an optional description, and searchable tags to keep your workspace organized. You'll also write top-level instructions that get sent to the AI whenever it interacts with this server. Keep these instructions clear and concise — they set the context for how the AI should behave across all of your tools.

Creating Tools for Your Server

Tools are the heart of your MCP server. Each tool is essentially a function stack designed to perform one specific action. When the AI receives a query, it reviews your full list of tools and determines which ones it needs to call to complete the task.

Here's how the example tools are structured:

• Get Basic User Info — Accepts a name or email input and queries your Xano user table to return matching records. You can customize the output to exclude fields you don't need.
• Check Admin Status — Takes a user ID and returns a simple true/false value indicating whether that user is an admin.
• Get Dog Breed Info — Accepts a breed ID and makes an external API request to a third-party dog breed API, returning relevant breed details.

Each tool also gets its own instructions. These should describe what the tool does and what inputs it expects. For more complex tools, consider including sample inputs and outputs to help the AI use them correctly.

Connecting to an AI Client

Once your tools are ready, it's time to connect your MCP server to a client. Using Claude Desktop as the example, you'll edit the claude_desktop_config.json file and paste in the standardized MCP server configuration from Xano's documentation. You'll need two things from Xano: your server's connection URL and an authentication token. The token controls access to your tools — it doesn't determine which user data is returned. After saving the config and restarting Claude, your tools will appear and be ready to use.

The Real Power of MCP: Taking Action

Retrieving data is just the beginning. The true value of MCP comes from enabling AI to take actions on behalf of your users. In the extended example, a fourth tool called Send Email is added. It triggers an API request to an email service and sends a compiled summary of user info — including admin status and dog breed — all from a single natural language prompt.

With Xano's MCP support, you can build AI-native functionality using the same visual builder you already know, give users natural language access to your data, and extend your application's capabilities in ways that weren't previously possible without deep AI integration.