Quest 7: Create an AI Agent with Tools from an MCP Server

The JS AI Build-a-thon is in full swing — and we’re turning up the power in Quest 7!
If you're just joining us, this is part of an ongoing challenge to help JavaScript and TypeScript developers build AI-powered apps from scratch. Catch up here and join the conversation on Discord.

Last quest, you dipped your toes into agentic development — giving your AI the ability to act and reason.
This time, we’re taking it further. In Quest 7, you’ll explore the Model Context Protocol (MCP) — a growing protocol in Agentic development that unlocks standardized tool-usage in AI agents via an MCP server.

Available tools for the OS Patrol agent

🎯 What You’ll Build

This quest is all about connecting your AI agent to tools that do real things.
You’ll:

• Create and spin up an MCP server using the MCP TypeScript SDK with tools that give your agent insights about your local OS (yes, your agent can now snoop your system — responsibly!)
• Connect it to your agent using the AI Toolkit in VS Code
• Learn how MCP lets you extend your agent’s abilities safely

In short? You’re about to build an agent with real-world reach following emerging protocols and industry practices.

🧰 What You’ll Need

Before you dive in, here’s what your toolbox should include:

• A GitHub Account
• Visual Studio Code installed
• Node.js installed
• The AI Toolkit extension for VS Code

⚠️ Before You Begin…

We strongly recommend completing Quest 6 (Build an AI Agent with AI Foundry) first. It lays down the essentials of creating and testing agents — giving you the perfect springboard for Quest 7.

🔌 Key Concepts You’ll Explore

Tool-Compatible Models

In this quest, you’ll create an agent that can access a tool to retrieve details about your system's OS. This introduces you to tool calling, a technique that’s shaping the next generation of powerful AI agents.

Tool-use models

Model Context Protocol (MCP)

MCP is an emerging standard for giving LLMs access to external tools in a structured, secure way. Instead of hardcoding every new function/ implementation, MCP allows you to define tools through a consistent standard — and your agent figures out when and how to use them. You’ll implement your first MCP server using the official TypeScript SDK.

Create a new MCP Server

AI Toolkit Integration

You’ll test and iterate your tool-augmented agent directly inside VS Code using the AI Toolkit extension. It provides a smooth playground for authoring, evaluating, and connecting to MCP tools.

Weather MCP Server - template

📚 Extra Resources

Want to go deeper? These resources will help you sharpen your understanding and build more advanced agent flows:

• MCP TypeScript SDK
  Build MCP-compatible tools and servers using TypeScript — straight from the source.
• Build Agents and Prompts in AI Toolkit
  Explore how to author agents, prompts, and test them with the AI Toolkit.

✅ Time to Build

This quest unlocks a whole new level of power for your agent. You’re not just building with AI — you’re giving AI the tools to build with you.

⏮️ Revisit the previous quest (Quest 6)