Multimodal, Real-Time AI Agent Systems

Chapter 10. Implementing MCP-Enabled Agents and Tool Servers

A Note for Early Release Readers

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited content as they write—so you can take advantage of these technologies long before the official release of these titles.

This will be the 11th chapter of the final book. Please note that the GitHub repo will be made active later on.

If you’d like to be actively involved in reviewing and commenting on this draft, please reach out to the editor at [e-mail].

In Chapter 9, we established the theoretical foundation of the MCP, a universal standard that decouples the intelligence of an agent from the tools and data it uses. We explored the architecture of Hosts, Clients, and Servers, and we discussed the primitives like Tools, Resources, and Prompts.

Now, it is time to stop talking about the protocol and start building with it.

In this chapter, we will move from architecture diagrams to running code. We are going to build a complete, end-to-end MCP system. We will see exactly how an agent built with the Google Agent Development Kit (ADK) reaches out, shakes hands with an external process, and acquires new capabilities dynamically at runtime.

The Scenario: The Financial Analyst

To demonstrate the power of this architecture, we will build a Financial Analyst Agent.

If we were building this the “old way” (as discussed in previous chapters), we would likely import a finance library directly ...