Build AI Agents With MCP

Tools and Installations

If you're on MacOS please use brew to install the packages

Claude free version has daily limits don't exhaust it before the event.

Once installed you should be able to run node, nvm,npx, uv, python commands from a terminal or shell and get some response. If you get command not found then the installation didn't happen properly

1. Install nodejs and npx.
2. Install python latest version (3.11+) and uv. to install uv.
3. Download and Install Claude Desktop
4. Clone this repo and run the following commands

uv venv
uv sync

---

Brief History of AI & Agents

1. What are LLMs?
2. What are agents, why we need them how we got here?
3. MCP

Step 1: Getting a Taste of MCP

Claude free version has daily limits don't exhaust it before the event.

Install Claude Desktop and get a taste of connecting the server to it - https://modelcontextprotocol.io/quickstart/user

Step 2: Build Your Own MCP Server

Building a weather agent server https://modelcontextprotocol.io/quickstart/server

Step 3: (Optional) Build Your Own MCP Client

Needs paid anthropic account.

If you're integrating Agentic systems into your own UI or other apps eg., Slack/Whatsapp/Telegram you will need the respective API Key from the model providers(OpenAI, Anthropic, Google)

We have two options here

Anthropic MCP Client Paid 5$ https://modelcontextprotocol.io/quickstart/client

Step 3: Agent SDK from Google vs MCP

https://cloud.google.com/vertex-ai/generative-ai/docs/agent-engine/develop/adk