Here’s a complete README.md you can drop into your repo AI-AGENTS-WITH-MCP. I’ve woven in the uv setup, prerequisites (Python 3.12+), and structured it around your three scenarios + Streamlit app.

---

AI-AGENTS-WITH-MCP

Building AI Agents with Model Context Protocol (MCP) and Amazon Bedrock

This repository contains a set of end-to-end examples demonstrating how to build decoupled, agentic AI systems using:

• Amazon Bedrock (Anthropic Claude models)
• Model Context Protocol (MCP) for tool decoupling
• Strands framework for agent orchestration
• uv for fast Python environment + dependency management
• Streamlit for a real-world demo with Zerodha Kite MCP

It follows the concepts from the article “Building AI Agents with MCP and Amazon Bedrock: An Implementation Guide” and showcases:

1. Baseline agents with Bedrock models
2. Local tool integration
3. Single MCP server connection
4. Multi-MCP orchestration
5. Custom MCP server (Calculator example)
6. Streamlit Kite portfolio assistant

---

🚀 Prerequisites

• Python 3.12+ installed on your system
• uv (fast Python package installer & runner)

Install uv:

pip install uv

Verify:

uv --version

---

🛠️ Setup Instructions

Clone the repo:

git clone https://github.com/<your-username>/AI-AGENTS-WITH-MCP.git
cd AI-AGENTS-WITH-MCP

Create a Python 3.12 virtual environment with uv:

uv venv --python 3.12

Activate the environment:

Linux/macOS

source .venv/bin/activate

Windows (PowerShell)

.venv\Scripts\activate

Install dependencies:

uv pip install -r requirements.txt

---

📂 Project Structure

AI-AGENTS-WITH-MCP/
├── README.md
├── requirements.txt
├── scenario1_single_server/   # Scenario 1: Baseline agent + local tool + doc MCP server
│   ├── baseline_agent.py
│   ├── agent_with_local_tool.py
│   └── agent_with_doc_mcp.py
├── scenario2_multi_server/    # Scenario 2: Multi-MCP orchestration
│   └── multi_server_agent.py
├── scenario3_custom_server/   # Scenario 3: Custom MCP server
│   ├── calculator_server.py
│   └── calculator_client.py
├── kite_streamlit_app/        # Real-world Streamlit + Kite MCP demo
│   └── streamlit_app.py
└── utils/                     # Shared helpers
    └── streamlit_helpers.py

---

📖 Scenarios

1. Single MCP Server

• Baseline agent with Claude (Bedrock)
• Add a local tool (e.g., Python execution)
• Connect to AWS Documentation MCP server (awslabs.aws-documentation-mcp-server)

Run:

python scenario1_single_server/agent_with_doc_mcp.py

---

2. Multi-MCP Orchestration

Agent with access to AWS Documentation + AWS Pricing MCP servers. Use case: Generate SageMaker fine-tuning research report.

Run:

python scenario2_multi_server/multi_server_agent.py

---

3. Custom MCP Server

Example of building your own MCP server (Calculator):

Start server:

python scenario3_custom_server/calculator_server.py

Run client:

python scenario3_custom_server/calculator_client.py

---

4. Streamlit Kite Portfolio Assistant

Interactive app integrating with Kite MCP and OpenAI GPT-4o.

Run:

streamlit run kite_streamlit_app/streamlit_app.py

Enter your OpenAI API key in the sidebar, log in to Zerodha Kite when prompted, and start asking portfolio questions.

📸 Example UI:

---

⚡ Using uvx for MCP servers

MCP servers are launched using uvx. Example:

uvx --from awslabs.aws-documentation-mcp-server@latest awslabs.aws-documentation-mcp-server.exe

Your agent code uses this under the hood.

---

📚 Learn More

• Model Context Protocol (MCP)
• Amazon Bedrock
• Strands

---

✍️ Author & Articles

This repo accompanies the Medium article: ➡️ “Building AI Agents with MCP and Amazon Bedrock: An Implementation Guide”

Follow-up post on LinkedIn coming soon 🚀

---

📜 License

MIT License. Use freely with attribution.