Recipe Genie - Real-time Data Integration with Amazon Bedrock Agents and MCP

An AI-powered recipe chatbot that demonstrates real-time data integration using Amazon Bedrock Agents, Strands SDK, and Model Context Protocol (MCP).

June 20, 2025 medium

bedrock python demo strands mcp agents real-time-data fastapi react

Recipe Genie - Real-time Data Integration with Amazon Bedrock Agents and MCP

An AI-powered recipe chatbot that demonstrates real-time data integration using Amazon Bedrock Agents, Strands SDK, and Model Context Protocol (MCP).

Overview

Recipe Genie showcases how to build intelligent AI agents that can access real-time external data sources through MCP servers. This demo creates a recipe assistant that:

Takes user ingredients as input and generates creative recipes
Integrates with real-time pricing data through an MCP server
Uses Amazon Bedrock with Claude 3.5 Sonnet for natural language understanding
Demonstrates the Strands Agents SDK for simplified agent development

This architecture pattern is valuable for building AI applications that need to combine language models with real-time data access, external APIs, or dynamic information sources. The MCP protocol provides a standardized way to expose tools and data to AI agents, making it easier to build modular, extensible AI systems.

Webserver UI

Technologies

Python 3.9+
AWS SDK (boto3)
Amazon Bedrock
Strands Agents SDK
Model Context Protocol (MCP)
FastAPI
React
WebSockets
Tailwind CSS

Difficulty

Medium

Prerequisites

Python 3.9+
Node.js 18+
AWS Account with Amazon Bedrock access
AWS credentials configured
API key for pricing data (see Setup section)

Enable AWS IAM permissions for Bedrock

The AWS identity you assume from your environment (which is the Studio/notebook Execution Role from SageMaker, or could be a role or IAM User for self-managed notebooks or other use-cases), must have sufficient AWS IAM permissions to call the Amazon Bedrock service.

To grant Bedrock access to your identity, you can:

Open the AWS IAM Console
Find your Role (if using SageMaker or otherwise assuming an IAM Role), or else User
Select Add Permissions > Create Inline Policy to attach new inline permissions, open the JSON editor and paste in the below example policy:

{
    "Version": "2012-10-17",
    "Statement": [
        {
  "Sid": "BedrockFullAccess",
  "Effect": "Allow",
  "Action": ["bedrock:*"],
  "Resource": "*"
        }
    ]
}

Enabling a Model in Amazon Bedrock Console

Follow the steps below to enable a foundation model for use in Amazon Bedrock:

Go to https://console.aws.amazon.com/
Select the desired AWS Region on the right top corner, e.g., us-east-1.
Navigate to Amazon Bedrock service.

Step 2: Go to the “Model access” section

In the left-hand navigation pane, click on Model access.
You will see a list of available foundation models from various providers (e.g., Anthropic, AI21, Meta, etc.).

Step 3: Request access to a model

Find the model you want to enable.
- For this demo we are using Claude 3.5 Sonnet v2, however, feel free to try different models. If you change the model, update the model id in your .env file.
Click Edit access next to the model.
Tick the checkbox to enable access to the model.

Step 4: Confirm and save

Click Save changes.
The model will now appear as available in your account, and you can start invoking it via the console or API.

Note: Some models may require approval or have usage limits based on your account setup.

For more information, refer to the Amazon Bedrock documentation.

Setup

1. Clone and Install

# Clone the repository
git clone https://github.com/aws-samples/sample-ai-possibilities.git
cd demos/recipe-genie-chatbot

# Install Python dependencies
pip install -r requirements.txt

# Install frontend dependencies
cd recipe-genie-frontend
npm install
cd ..

2. Configuration

The project uses two .env files - one for the backend services and one for the frontend:

Backend Configuration (root directory)

# Copy and configure backend environment
cp .env.example .env

Frontend Configuration

# Copy and configure frontend environment
cd recipe-genie-frontend
cp .env.example .env
cd ..

Edit .env files with your configuration:

⚠️ IMPORTANT SECURITY NOTICE:

NEVER commit credentials to version control
NEVER share your AWS access keys publicly
The .env file should be added to .gitignore
For EC2 deployments, use IAM roles instead of access keys (see below)

Required configuration in root .env:

COLES_API_KEY: API key for pricing service
AWS_REGION: Your AWS region (default: us-east-1)
AWS_ACCESS_KEY_ID: Your AWS access key (local development only)
AWS_SECRET_ACCESS_KEY: Your AWS secret key (local development only)

Optional configuration in root .env:

BEDROCK_MODEL_ID: Bedrock model to use (default: us.anthropic.claude-3-5-sonnet-20241022-v2:0)
STORE_ID: Default store ID for pricing (default: 0584)
MCP_HOST: MCP server host (default: localhost)
MCP_PORT: MCP server port (default: 8000)
API_HOST: API server host (default: 0.0.0.0)
API_PORT: API server port (default: 8080)

Required configuration in recipe-genie-frontend/.env:

REACT_APP_API_URL: Backend API URL (default: http://localhost:8080)

Recommended: Using IAM Roles on EC2

For EC2 deployments, we strongly recommend using IAM roles instead of hardcoded credentials:

Create an IAM role with Bedrock access permissions
Attach the role to your EC2 instance
Remove AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY from .env
The SDK will automatically use the instance role

This approach is more secure and follows AWS best practices.

3. Start the Services

Run all three components in separate terminals:

# Terminal 1: Start the MCP server
python coles_real_api_mcp.py

# Terminal 2: Start the backend API
python recipe_genie_api.py

# Terminal 3: Start the frontend
cd recipe-genie-frontend
npm start

The application will be available at http://localhost:3003

Architecture

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│   React     │────▶│   FastAPI    │────▶│  MCP Server │
│  Frontend   │     │   Backend    │     │  (Pricing)  │
└─────────────┘     └──────────────┘     └─────────────┘
                  │
                  ▼
          ┌────────────────────┐
          │ Amazon Bedrock     │
          │   Claude 3.5       │
          └────────────────────┘

Key Features

Real-time Streaming: WebSocket connection for streaming AI responses
MCP Integration: Demonstrates tool usage through the MCP protocol
Error Handling: Graceful fallback from WebSocket to HTTP
Session Management: Maintains conversation history
Responsive UI: Mobile-friendly React interface

Project Structure

recipe-genie-chatbot/
├── coles_real_api_mcp.py    # MCP server for pricing data
├── recipe_genie_api.py       # FastAPI backend with Strands
├── requirements.txt          # Python dependencies
├── .env.example             # Backend environment template
├── .gitignore               # Git ignore file
└── recipe-genie-frontend/   # React frontend
    ├── src/
    │   └── components/
    │       └── RecipeGenieChat.js
    ├── package.json
    └── .env.example         # Frontend environment template

Deployment Considerations

For production deployment:

Security:
- Always use AWS IAM roles instead of access keys in production
- Never store credentials in code or configuration files
- Implement proper CORS policies (remove wildcard origins)
- Add rate limiting and input validation
- Use secrets management services (AWS Secrets Manager, Parameter Store)
Scalability:
- Use Redis or DynamoDB for session storage
- Deploy with container orchestration (ECS/EKS)
- Implement proper logging and monitoring
Performance:
- Add caching for common queries
- Use CDN for frontend assets
- Optimize WebSocket connections

Cleanup Instructions

To remove all resources:

Stop all running processes (Ctrl+C in each terminal)

Remove installed dependencies:

# Remove Python virtual environment if used
deactivate
rm -rf venv/
   
# Remove node modules
cd recipe-genie-frontend
rm -rf node_modules/

Delete the project directory

Notes

This project is for educational purposes and is not designed for production use. For production use, please implement proper security, scalability, and compliance measures appropriate for your use case.
The MCP server implementation is a demonstration of the protocol and should be adapted to your specific data sources
Ensure you comply with all API terms of service and data usage policies

Security

See CONTRIBUTING for more information.

License

This library is licensed under the MIT-0 License. See the LICENSE file.

Full Source Code

View on GitHub

Recipe Genie - Real-time Data Integration with Amazon Bedrock Agents and MCP

Overview

Tags

Technologies

Difficulty

Prerequisites

Enable AWS IAM permissions for Bedrock

Enabling a Model in Amazon Bedrock Console

Step 1: Sign in to the AWS Console

Step 2: Go to the “Model access” section

Step 3: Request access to a model

Step 4: Confirm and save

Setup

1. Clone and Install

2. Configuration

Backend Configuration (root directory)

Frontend Configuration

Recommended: Using IAM Roles on EC2

3. Start the Services

Architecture

Key Features

Project Structure

Deployment Considerations

Cleanup Instructions

Notes

Security

License

Full Source Code