MCP AWS re:Invent Demo with AWS Bedrock LLM Integration

A comprehensive demonstration of the Model Context Protocol (MCP) featuring AWS Bedrock LLM as an intelligent agent that uses MCP servers as tools. Built for AWS re:Invent 2025 conference planning with real data and live API integrations.

🧠 Strands + MCP Architecture

┌─────────────────────────────────────────────────────────────────────────────┐
│                           MCP Client                                        │
│  ┌─────────────────┐    ┌─────────────────────────────────────────────────┐ │
│  │  React Frontend │    │        Strands Agent               │ │
│  │  (Chat UI)      │◄──►│  ┌─────────────────┐  ┌─────────────────────┐  │ │
│  └─────────────────┘    │  │ Strands Agent   │  │   Fallback MCP      │  │ │
│                         │  │        │  │   Tool Execution    │  │ │
│                         │  └─────────────────┘  └─────────────────────┘  │ │
│                         └─────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────┘
                                            │
                              ┌─────────────┴─────────────┐
                              │     MCP Protocol          │
                              │     (JSON-RPC)            │
                              └─────────────┬─────────────┘
                                            │
        ┌───────────────────────────────────┼───────────────────────────────────┐
        │                                   │                                   │
┌───────▼────────┐   ┌────────▼────────┐   ┌───────▼────────┐   ┌───────▼────────┐
│ re:Invent      │   │ Weather Server  │   │ Stock Server   │   │ AWS Docs       │
│ MCP Server     │   │ (Live APIs)     │   │ (Live APIs)    │   │ MCP Server     │
│ (Local)        │   │ (Local)         │   │ (Local)        │   │ (Remote)       │
│                │   │                 │   │                │   │                │
│ Strands Tools: │   │ Strands Tools:  │   │ Strands Tools: │   │ Strands Tools: │
│ • searchSessions│   │ • getWeather    │   │ • getStock     │   │ • searchAWS    │
│ • recommend    │   │ • getForecast   │   │ • getMarket    │   │ • getGuides    │
│ • getDetails   │   │                 │   │                │   │                │
│                │   │ Resources:      │   │ Resources:     │   │ Resources:     │
│ Resources:     │   │ • wttr.in       │   │ • Yahoo Finance│   │ • Official     │
│ • SQLite DB    │   │ • OpenWeather   │   │ • Alpha Vantage│   │   AWS Docs     │
│ • 290 Sessions │   │                 │   │                │   │ • uvx runtime  │
└────────────────┘   └─────────────────┘   └────────────────┘   └────────────────┘

🚀 Key Features

Strands Agent Architecture

AWS Bedrock Agent with Strands for intelligent tool orchestration
Graceful Fallback to direct MCP tool execution when agent setup is complex
Claude 3 Haiku as the reasoning engine
Automatically selects appropriate MCP tools based on user queries
Natural language understanding and contextual responses
Multi-tool orchestration for complex requests

Real Conference Data

2254 actual AWS re:Invent 2025 sessions parsed from official conference catalog
Full-text search across session titles, descriptions, and speakers
Smart recommendations based on interests and experience level
Complete session details with times, locations, and tracks

Live Data Integration

Weather: Real-time data from wttr.in API with OpenWeatherMap fallback
Stock Market: Live prices from Yahoo Finance API
Travel: Mock flight data for Las Vegas routes (LAS airport)

MCP Architecture

Tools: Functions that MCP servers expose to the LLM
Resources: Data sources (databases, APIs, files)
Prompt Templates: Structured prompts for consistent interactions

🛠️ Setup & Installation

Prerequisites

Node.js 18+ and npm
AWS Account with Bedrock access
AWS CLI installed
Python and uv package manager (for remote MCP servers)

Quick Start

# 1. Install dependencies
npm install

# 2. Install uv (Python package manager for remote MCP servers)
# macOS/Linux:
curl -LsSf https://astral.sh/uv/install.sh | sh
# Or via pip: pip install uv

# 3. Configure AWS credentials (recommended)
aws configure
# Enter your AWS Access Key ID, Secret Access Key, and region (us-east-1)

# 4. Optional: Create .env file for custom configuration
cp .env.example .env
# Edit .env with AWS_REGION or API keys if needed

# 5. Start the demo
chmod +x start-llm-demo.sh
./start-llm-demo.sh

# OR start manually:
npm start                    # Backend with strands agent
npm run start:demo          # Both backend and frontend

# Note: The SQLite database with 2254 re:Invent sessions will be automatically 
# created in servers/reinvent_schedule.db on first run

AWS Configuration Options

Option 1: AWS CLI (Recommended)

aws configure
# AWS Access Key ID: your_key_here
# AWS Secret Access Key: your_secret_here  
# Default region: us-east-1
# Default output format: json

Option 2: Environment Variables

export AWS_ACCESS_KEY_ID=your_key_here
export AWS_SECRET_ACCESS_KEY=your_secret_here
export AWS_REGION=us-east-1

Option 3: .env File (Optional)

cp .env.example .env
# Edit .env with AWS_REGION (credentials optional if using aws configure)
# Optionally add API keys for enhanced live data:
# WEATHER_API_KEY=your_openweathermap_key
# ALPHA_VANTAGE_API_KEY=your_alpha_vantage_key

🔧 MCP Tools & Resources

AWS re:Invent Schedule MCP Server

Tools:

search_sessions - Find sessions by keywords, topics, speakers
get_session_details - Get complete session information
recommend_sessions - AI-powered personalized recommendations

Resources:

SQLite database with 2254 real AWS re:Invent sessions
Full-text search capabilities
Session metadata (tracks, levels, days, locations)

Weather MCP Server

Tools:

get_weather - Current weather conditions
get_forecast - 3-day weather forecast

Resources:

wttr.in API (primary, no key required)
OpenWeatherMap API (fallback)

Stock MCP Server

Tools:

get_stock_price - Real-time stock quotes
get_market_status - Trading hours and market state

Resources:

Yahoo Finance API (primary, no key required)
Alpha Vantage API (fallback)

Flight MCP Server

Tools:

search_flights - Find flights to Las Vegas for re:Invent

Resources:

Mock flight database with realistic data
Las Vegas (LAS) focused routes

AWS Documentation MCP Server (Remote)

Tools:

search_docs - Search AWS service documentation using official AWS MCP server
Powered by uvx awslabs.aws-documentation-mcp-server@latest

Resources:

Official AWS service documentation
Real-time access to AWS docs via remote MCP server
Comprehensive coverage of all AWS services

🎯 Demo Experience

Natural Language Queries

The LLM agent understands complex, conversational requests:

"I'm interested in AI and serverless. What re:Invent sessions would you recommend, 
and what's the weather like in Las Vegas?"

The agent automatically:

Calls recommend_sessions with interests: ["AI", "leadership"]
Calls get_weather with location: "Las Vegas"
Combines results in a natural response

Sample Interactions

Session Search: "Show me cybersecurity sessions on Day 2"
Weather: "What's the weather forecast for Las Vegas?"
Travel: "Find flights from San Francisco to Las Vegas"
Recommendations: "Suggest sessions for a beginner in data science"
AWS Documentation: "How do I use Bedrock with Lambda functions?"
MCP Protocol: "What are MCP tools and how do they work?"
Complex: "Plan my re:Invent trip: weather, flights from NYC, AI sessions, and AWS Bedrock documentation"

🏗️ Technical Architecture

Database Setup & Schema

Automatic Initialization: The SQLite database (servers/reinvent_schedule.db) is automatically created and populated when you first run the server. It contains:

2254 real AWS re:Invent 2025 sessions parsed from the official conference catalog
Full session metadata: titles, speakers, descriptions, times, locations, tracks
Searchable fields: All text fields are indexed for full-text search

Database Schema:

sessions table: 2254 real AWS re:Invent sessions with complete metadata
Indexed fields: title, description, speaker, tags, track, level, day
Search capabilities: Full-text search with relevance scoring across all fields

Verify Database:

# Check if database exists and has data
sqlite3 servers/reinvent_schedule.db "SELECT COUNT(*) FROM sessions;"
# Should return: 2254

# View sample sessions
sqlite3 servers/reinvent_schedule.db "SELECT title, type, level FROM sessions LIMIT 3;"

Database Recreation (if needed): If the database file is missing or corrupted, it will be automatically recreated with sample data when the server starts. The full 290-session dataset is embedded in the application and loaded automatically.

Strands Agent Integration

Primary: AWS Bedrock Agent with strands architecture
Fallback: Direct MCP tool execution with Claude 3 Haiku
Tool Planning: Intelligent selection of appropriate MCP tools
Parameter Extraction: Smart parsing of user queries
Response Generation: Natural language formatting of results
Graceful Degradation: Always functional regardless of AWS setup complexity

Fallback Strategy

LLM failure → Rule-based query processing
API failures → Simulated data with clear indicators
Graceful degradation ensures demo always works

📊 Demo Script

See DEMO_SCRIPT.md for a complete 5-minute walkthrough including:

Setup instructions
Key talking points
Expected responses
Troubleshooting tips
Architecture explanations

This demo showcases the future of AI applications: Strands agents using MCP as a standard protocol to intelligently orchestrate multiple data sources and services.

🧪 Testing the Demo

Test AWS Documentation Integration

# Test the remote AWS MCP server directly
curl -X POST "http://localhost:3001/api/chat" \
  -H "Content-Type: application/json" \
  -d '{"message": "How do I get started with AWS Bedrock?"}'

# Expected: Real AWS documentation with official links

Test AWS re:Invent Conference Data

# Test the local SQLite database
curl -X POST "http://localhost:3001/api/chat" \
  -H "Content-Type: application/json" \
  -d '{"message": "Show me AI sessions at re:Invent"}'

# Expected: Real conference sessions from 2254-session database

Test Live Data Integration

# Test weather API
curl -X POST "http://localhost:3001/api/chat" \
  -H "Content-Type: application/json" \
  -d '{"message": "What is the weather in Seattle?"}'

# Expected: Live weather data from wttr.in

Test Multi-Tool Orchestration

# Test complex query using multiple MCP servers
curl -X POST "http://localhost:3001/api/chat" \
  -H "Content-Type: application/json" \
  -d '{"message": "Plan my re:Invent trip: weather in Las Vegas, AI sessions, and AWS Bedrock documentation"}'

# Expected: Combined results from weather, sessions, and AWS docs

Verify System Status

# Check overall system health including database
curl "http://localhost:3001/api/health"

# Look for:
# - "database": "connected"
# - "sessionsCount": 290
# - "llmAgent": "enabled"

# Check database info specifically
curl "http://localhost:3001/api/database/info"

# Should show:
# - totalSessions: 290
# - Available tracks, levels, and days

Frontend Testing

Open http://localhost:3000
Click "MCP Architecture" tab to see strands architecture
Test queries in chat interface:
- "How do I use Bedrock with Lambda?" (AWS docs)
- "Show me security sessions" (re:Invent data)
- "Weather in Las Vegas" (live data)

Troubleshooting

AWS docs not working: Check uvx installation (uvx --version)
Agent working normally: The system uses direct MCP tool orchestration
No responses: Verify AWS credentials (aws sts get-caller-identity)

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
frontend		frontend
servers		servers
DEMO_SCRIPT.md		DEMO_SCRIPT.md
README.md		README.md
package-lock.json		package-lock.json
package.json		package.json
reinvent-2025-sessions.csv		reinvent-2025-sessions.csv
start-llm-demo.sh		start-llm-demo.sh

neelamkoshiya/Reinvent_Assistant

Folders and files

Latest commit

History

Repository files navigation

MCP AWS re:Invent Demo with AWS Bedrock LLM Integration

🧠 Strands + MCP Architecture

🚀 Key Features

Strands Agent Architecture

Real Conference Data

Live Data Integration

MCP Architecture

🛠️ Setup & Installation

Prerequisites

Quick Start

AWS Configuration Options

🔧 MCP Tools & Resources

AWS re:Invent Schedule MCP Server

Weather MCP Server

Stock MCP Server

Flight MCP Server

AWS Documentation MCP Server (Remote)

🎯 Demo Experience

Natural Language Queries

Sample Interactions

🏗️ Technical Architecture

Database Setup & Schema

Strands Agent Integration

Fallback Strategy

📊 Demo Script

🧪 Testing the Demo

Test AWS Documentation Integration

Test AWS re:Invent Conference Data

Test Live Data Integration

Test Multi-Tool Orchestration

Verify System Status

Frontend Testing

Troubleshooting

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages