Teach on Udemy

Turn what you know into an opportunity and reach millions around the world.

Learn More

Your cart is empty.

Keep shopping

MCP & A2A - Model Context Protocol & Agent to Agent Protocol

Name: MCP & A2A - Model Context Protocol & Agent to Agent Protocol
Rating: 4.3 (142 reviews)

New 2026 Updates for MCP version 2 | Build 6 MCP Clients, 4 Servers, UI | Connect 3 Agents via A2A | FREE Gemini Key

Created byKartik Marwah

Last updated 8/2026

English

English [Auto],

What you'll learn

MCP 2026 updates added to course for MCP Server & Client
You will learn Agent to Agent (A2A) Protocol - Connect 3 agents via A2A and build a host orchestrator agent with A2A + MCP
You'll build 5 MCP Clients and 3 MCP Servers from Scratch. Full Working Code Included!
Build a UI for MCP client in Streamlit in Python
Learn to deploy your MCP Server to Google Cloud using Server Sent Events (SSE)
We'll use a free Gemini API Key, so you don't need to pay for AI Models when learning!
We'll use Python, Gemini, LangGraph, SSE, Streamlit on MacOS

Course content

28 sections • 231 lectures • 14h 45m total length

What you will learn in this course?7:55
Learn to design and build a complete multi-agent system using MCP and the two way protocol, from basic concepts to servers, clients, registries, and a host orchestrator.
What does MCP help us build?0:58
Why do we need MCP?1:43
MCP Overview and 4 Key Elements in MCP1:36
Can you answer these MCP basics?
New Upcoming MCP Version 2 2026 updates0:52

Install Python, UV, VS Code and Git0:14
MacOS - Install Python, UV, VS Code, Git5:00
Set up your environment for building your first agent by installing Python 3.13.4, uv, VS Code, and Git on macOS, Windows, and Ubuntu, then verify installations.
Windows - Install Python, UV, VS Code, Git7:15
Ubuntu - Install Python, UV, VS Code, Git4:42
Set up Ubuntu for Python development by installing python3, pip, and venv with apt, install UB via curl, then install VS Code and the Python extension, and finally install git.
Resources - Get Code on Github2:59
Resources - Class Github Repositories List0:18
Resources - Official Links0:16

Introduction1:44
Build an mcp server with the new mcp server class as model context protocol moves to v2. Get upgrade guidance and use the mcp builder with Gemini API key.
MCP Version 2 - Breaking Changes to FastMCP1:38
Overview of the Terminal MCP Server We’ll Build1:32
Getting the code0:18
Accessing the Online MCP Server Builder with Code0:28
Directory Structure Walkthrough2:03
This directory structure walkthrough explains the terminal server layout, including main.py, tools.py, and a workspace, plus environment and cloud desktop config files, attribution, license, readme, and requirements.
Code Walkthrough 1 - Server Definition in main.py file3:09
Code Walkthrough 2 - Tools Code in tools.py file4:55
Code Walkthrough 3 - claude_desktop_config.json file2:29
Code Walkthrough 4 - .env file0:25
Configure the server environment with a .env file and workspace path to enable running with or without cloud desktop config.json, and prepare the codebase for execution.
Export the code to your own Github (optional)1:31
Clone the repo code1:05
Install Claude for Dektop1:11
We will use Claude for Desktop Application as our MCP Client. So we install it here.
Install Claude Desktop (for Windows Users Only)0:46
Setup claude_desktop_config.json file3:40
Connect Your Server To Claude Desktop (for Windows users only)0:59
Important - Project Dependencies & Set Workspace [for both MacOS + Windows]0:45
Testing the MCP Server with Claude Desktop2:10
Demonstrate testing the MCP server with Claude Desktop by using the terminal server to run shell commands that create mcpsuccess.txt, illustrating stdio transport version 2 and the free learning plan.
Conclusion0:59
Build and run the MCP server using protocol v2 with stdio transport, implement logging and execute command, and integrate with cloud desktop to execute commands and create files.

Introduction1:24
Explore upgrading from model context protocol v1 to v2, build MCP clients and servers with a terminal setup, and connect via the Google agent development kit using stdio transport.
Quick Demo - What We’ll Build3:00
Getting the Code0:29
Getting the Code0:18
Code Setup2:24
Initialize uv project, create and activate a virtual environment, install requirements, open the 02mcp-adk-client folder in visual studio code, and configure .env with google api key to start two servers.
MCP Client Code Walkthrough - manager.py3:48
Explore the MCP client through the manager.py walkthrough, showing json rpc over stdio, async exit stack lifecycle management, config loading, tool discovery, multiplexing, and shutdown.
MCP Client - Server Discovery from Config1:34
Load the MCP server configurations from config.json using loadConfig, extract each server's name and details, and build stdio server parameters from command arguments and environment.
MCP Client - Server Connection Logic3:13
MCP Client - Listing All Tools1:12
MCP Client - Calling a Tool1:42
MCP Client - Session Cleanup1:04
MCP Server - Simple Echo Server Example1:43
Running the MCP Client Manager0:59

ADK + MCP Integration - Overview and Step 1 (Load Server Config)3:06
Create an autonomous llm agent using Google adk to discover and use tools from external mcp servers, loading server configuration via config.json and environment variables.
ADK Integration - Step 2 (Connect to MCP Servers)3:07
Connect to MCP servers using the Google ADK tool sets to manage stdio connections, discover tools with toolset.getTools(), and display a name and description table while handling exceptions.
ADK Integration - Step 3 (Initialize Agent with McpToolset)0:50
ADK Integration - Step 4 (Initialize the Runner)0:42
Initialize the in-memory runner with the agent and app name to power the loop, taking user input, sending it to the agent, executing tools when called, and returning results.
ADK Integration - Step 5 (Start a Conversation Session)1:24
Start a conversation session with the runner to let the agent remember context across messages, and create a conversation session with app name and user ID.
ADK Integration - Step 6 (Interactive Chat Loop)4:40
MCP Client Manager vs ADK Command Line Integration2:26
Running the Code1:18

Introduction to Config.json for multiple server support0:36
Here I introduce the concept for multi-mcp server support for our MCP client
Let's look at the config.json file0:36
Instructions for Windows Users3:27
Demo - MCP Client with Multiple MCP Servers4:10
MCP Client (with json config) Code Walk Through - Part 12:23
Why Choose Gemini 2.0 Flash and not the Pro Models0:47
MCP Client (with json config) Code Walk Through - Part 2 (continued)3:52
How to use existing MCP servers from MCP Github (example uses "fetch" server)3:57
Clone MCP reference servers from the MCP GitHub and use the fetch server to fetch content. Build its Docker image, configure the MCP client with config.json, and run for testing.

Introduction to SSE and Preview of What We'll Build0:58
Quick Recap, What is STDIO and SSE?2:12
Setup Directories, Clone GitHub Code (git pull only, if done at course start)1:35
Set up a clean project structure by creating mcp directories for servers, clients, and workspace, then clone the server and client code from GitHub into terminal_server and mcpi_client directories.
Setup Virtual Environment and Dependencies2:27
Instructions for Windows Users2:36
MCP SSE Server Code Walkthrough8:57
MCP SSE Client Code Walkthrough14:10
Demonstrates building an MCP SSE client that connects to an MCP server, integrates the Gemini SDK for function calling, and handles tool calls and prompts.
Dockerfile Code (for MCP Server) Walkthrough0:55
Test your MCP SSE Server and Client Locally2:36

Create a new Gmail Account (if you prefer to)2:35
Disclaimer
The information provided in this course is for educational purposes only.
Please be advised that when creating a Google account, Google Cloud Platform (GCP) project, or any other third-party account or service, you are solely responsible for reviewing and understanding the applicable terms and conditions, privacy policies, pricing, usage limits, and any other relevant policies or charges associated with that service.
The instructor, course provider, and any affiliated parties do not provide any guarantees, warranties, or representations regarding the accuracy, completeness, or current applicability of any third-party services mentioned in this course, including but not limited to Google Cloud's $300 credit offer. Policies and offerings may change at any time without notice, and the information in this course may become outdated.
By proceeding to create and use any such account or service, you agree that you do so at your own discretion and risk. The instructor and related parties shall not be held liable for any losses, charges, damages, liabilities, or consequencesarising from your use or attempted use of such services.
Always exercise independent judgment and due diligence before engaging with any third-party platform or offer.
Create a Google Cloud Project2:48
Disclaimer
The information provided in this course is for educational purposes only.
Please be advised that when creating a Google account, Google Cloud Platform (GCP) project, or any other third-party account or service, you are solely responsible for reviewing and understanding the applicable terms and conditions, privacy policies, pricing, usage limits, and any other relevant policies or charges associated with that service.
The instructor, course provider, and any affiliated parties do not provide any guarantees, warranties, or representations regarding the accuracy, completeness, or current applicability of any third-party services mentioned in this course, including but not limited to Google Cloud's $300 credit offer. Policies and offerings may change at any time without notice, and the information in this course may become outdated.
By proceeding to create and use any such account or service, you agree that you do so at your own discretion and risk. The instructor and related parties shall not be held liable for any losses, charges, damages, liabilities, or consequencesarising from your use or attempted use of such services.
Always exercise independent judgment and due diligence before engaging with any third-party platform or offer.
Install and Setup Google Cloud Command Line Interface (gcloud CLI)5:33
Please create the development directory using
mkdir ~/development

Disclaimer
The information provided in this course is for educational purposes only.
Please be advised that when creating a Google account, Google Cloud Platform (GCP) project, or any other third-party account or service, you are solely responsible for reviewing and understanding the applicable terms and conditions, privacy policies, pricing, usage limits, and any other relevant policies or charges associated with that service.
The instructor, course provider, and any affiliated parties do not provide any guarantees, warranties, or representations regarding the accuracy, completeness, or current applicability of any third-party services mentioned in this course, including but not limited to Google Cloud's $300 credit offer. Policies and offerings may change at any time without notice, and the information in this course may become outdated.
By proceeding to create and use any such account or service, you agree that you do so at your own discretion and risk. The instructor and related parties shall not be held liable for any losses, charges, damages, liabilities, or consequencesarising from your use or attempted use of such services.
Always exercise independent judgment and due diligence before engaging with any third-party platform or offer.
Instructions for Windows Users4:37
Build Docker Image for Google Cloud1:02
Deploy MCP SSE Server to Google Cloud Run0:58
Disclaimer
The information provided in this course is for educational purposes only.
Please be advised that when creating a Google account, Google Cloud Platform (GCP) project, or any other third-party account or service, you are solely responsible for reviewing and understanding the applicable terms and conditions, privacy policies, pricing, usage limits, and any other relevant policies or charges associated with that service.
The instructor, course provider, and any affiliated parties do not provide any guarantees, warranties, or representations regarding the accuracy, completeness, or current applicability of any third-party services mentioned in this course, including but not limited to Google Cloud's $300 credit offer. Policies and offerings may change at any time without notice, and the information in this course may become outdated.
By proceeding to create and use any such account or service, you agree that you do so at your own discretion and risk. The instructor and related parties shall not be held liable for any losses, charges, damages, liabilities, or consequencesarising from your use or attempted use of such services.
Always exercise independent judgment and due diligence before engaging with any third-party platform or offer.
Test MCP SSE Server on Google Cloud2:04
Disclaimer
The information provided in this course is for educational purposes only.
Please be advised that when creating a Google account, Google Cloud Platform (GCP) project, or any other third-party account or service, you are solely responsible for reviewing and understanding the applicable terms and conditions, privacy policies, pricing, usage limits, and any other relevant policies or charges associated with that service.
The instructor, course provider, and any affiliated parties do not provide any guarantees, warranties, or representations regarding the accuracy, completeness, or current applicability of any third-party services mentioned in this course, including but not limited to Google Cloud's $300 credit offer. Policies and offerings may change at any time without notice, and the information in this course may become outdated.
By proceeding to create and use any such account or service, you agree that you do so at your own discretion and risk. The instructor and related parties shall not be held liable for any losses, charges, damages, liabilities, or consequencesarising from your use or attempted use of such services.
Always exercise independent judgment and due diligence before engaging with any third-party platform or offer.

Requirements

Code WalkThroughs Included, no in-depth experience needed
Basic Python is a must
Basic familiarity with LLMs like Google Gemini/Claude
Course Recorded using MacOS. Windows Users can follow at their discretion

Description

NEW (FEB 2026) - MCP Version 2 updates added to course for MCP Server. Note that other updates are in progress! MCP is migrating to Version 2 by around end of March 2026. I am adding updates to the course to reflect those changes.

TLDR - In this course, you'll learn about MCP & A2A - Model Context Protocol & Agent2Agent Protocol. We will build 5 MCP Clients and 3 MCP Servers from Scratch. You'll build a UI for MCP client in Streamlit in Python. You'll learn to deploy your MCP Server to Google Cloud using Server Sent Events. To top it all, we'll use a free Gemini API Key from Google so you don't need to pay for AI Models when learning! We then build A2A Agents, Clients and Servers and use Google's ADK or Agent Development Kit. Integrate MCP + A2A together

Key Objectives of Course

Use of MCP v2 (new update expected around end of March 2026)
MCP Authentication - Sign in to MCP server with Google!
Build A2A Servers, Clients and Agents (using Google's ADK or agent development kit) - in Beta
Connect 3 Agents via A2A with full code implementation
Build a Host Orchestrator agent - integrate with A2A + MCP
A2A Python SDK
Build 5 MCP Clients:
1. #1. Basic Python + Gemini
2. #2. LangGraph Based
3. #3. LangGraph (with config.json)
4. #4. Server Sent Events (SSE)
5. #5. Streamlit MCP Client
6. #6 Streamable HTTP Client
Build 4 MCP Servers:
1. #1 Python STDIO based
2. #2 Python STDIO (with Docker)
3. #3 Server Sent Event (SSE) Server
4. #4 Streamable HTTP Server
Connect your server to Claude Desktop
Deploy MCP Server on Google Cloud Platform. Yes, you heard that right :)
Build Streamlit UI for MCP Client in Python
Avoid AI Costs using a free Gemini API Key: to work with Gemini API, Tools and Function calling

Detailed Course Description

Learn basics of MCP

In this hands-on course, you will gain a solid foundation in the Model Context Protocol (MCP) and its bidirectional client-server architecture. We begin by exploring the fundamentals of MCP, equipping you with the conceptual understanding needed to work confidently with this powerful protocol.

Next, you will implement your own MCP server in Python, hosted on a MacBook. This server will be connected to Claude Desktop, showcasing how you can interact with external tools. As part of this quick start, you'll build and test a tool that executes terminal commands via MCP using Claude Desktop.

Building on this foundation, we’ll guide you through developing your own MCP client in Python, integrating it with Google’s Gemini API. With a free Gemini API Key from Google, you’ll connect your client to your MCP server and perform tool invocations just like Claude Desktop.

By the end of this course, you will:

Understand the basic architecture and flow of MCP
Learn to build and run an MCP server using Python
Connect your custom MCP server to Claude Desktop
Obtain and configure a free Google Gemini API Key
Develop your own Python-based MCP client powered by Google Gemini
Test the interaction between your client and server to run real commands

What This Course Does Not Cover:

Advanced MCP primitives beyond tool-calling (for now, we focus only on tool invocation)
Windows-specific setup: While many steps are applicable across platforms, this course uses a MacBook for demonstrations. Windows users may follow along at their discretion.

Important Notes:

This course uses tools, APIs, Keys and services provided by third-party companies and open source projects. We do not offer any warranties or guarantees related to these services. Learners are responsible for understanding and agreeing to the terms and policies of each provider.
While many resources exist in fragments across the web, this course brings everything together in a streamlined, tested, and learner-friendly format based on real-world implementations.
Disclaimer
The information provided in this course is for educational purposes only.
Please be advised that when creating a Google account, Google Cloud Platform (GCP) project, using the Gemini API Key or using any other third-party account or service, you are solely responsible for reviewing and understanding the applicable terms and conditions, privacy policies, pricing, usage limits, and any other relevant policies or charges associated with that service.
The instructor, course provider, and any affiliated parties do not provide any guarantees, warranties, or representations regarding the accuracy, completeness, or current applicability of any third-party services mentioned in this course, including but not limited to Google Cloud's $300 credit offer and the free Gemini API Key. Policies and offerings may change at any time without notice, and the information in this course may become outdated.
By proceeding to create and use any such account or service, you agree that you do so at your own discretion and risk. The instructor and related parties shall not be held liable for any losses, charges, damages, liabilities, or consequences arising from your use or attempted use of such services.
Always exercise independent judgment and due diligence before engaging with any third-party platform or offer. The credits and key from Google are available from Google irrespective of whether you take this course or not. We use them to help students setup and learn about MCP and AI without incurring cloud/AI costs.

Whether you're a developer, tinkerer, or AI enthusiast, this course will empower you to build practical systems using cutting-edge AI and modern protocol design.

Who this course is for:

Take this course if you want to code MCP Clients, Servers and User Interfaces
AI Agent Developers, Software Engineers, LangChain Developers
Entrepreneurs looking for a QuickStart on MCP

MCP & A2A - Model Context Protocol & Agent to Agent Protocol

What you'll learn

Explore related topics

Course content

Introduction5 lectures • 13min

Environment Setup (MacOS, Windows, Ubuntu) & Resources7 lectures • 21min

[MCP v2 2026 Update] Build Your Own MCP Server19 lectures • 32min

[MCP v2 2026 Update] Build your Own MCP Client Using Python + Google ADK: Part 113 lectures • 23min

[MCP v2 2026 Update] Build your Own MCP Client Using Python + Google ADK: Part 28 lectures • 18min

Build Docker MCP Server2 lectures • 18min

Build LangChain MCP Client2 lectures • 13min

Build MCP Client with Multiple Server Support8 lectures • 20min

Server Sent Events - MCP Server and Clients using SSE9 lectures • 36min

Deploying MCP Server to Google Cloud Platform7 lectures • 20min

Requirements

Description

Who this course is for: