Auto RFP - AI Agent Demo

An end-to-end AI agent system that demonstrates Context Engineering and A2A (Agent-to-Agent) communication for automated RFP processing and proposal generation.

🎯 Demo Overview

This system showcases advanced AI agent capabilities through a complete RFP workflow:

1. RFP Discovery - Bright Data MCP fetches fresh RFP listings from public portals
2. PDF Processing - Apify actor extracts and chunks RFP content
3. Context Engineering - Senso normalizes extracted data into structured schema
4. Retrieval Index - Redis VL + LlamaIndex build knowledge graph
5. Proposal Generation - Agent drafts proposal using retrieved context
6. A2A Evaluation - HoneyHive evaluates and agent self-improves
7. Microsite Deploy - Qodo publishes proposal microsite

🚀 Quick Start

# Clone and setup
git clone <repository-url>
cd auto-rfp

# Install dependencies
npm install

# Copy environment template
cp .env.example .env.local

# Start development server
npm run dev

# Open http://localhost:3000

⚙️ Setup Instructions

Prerequisites

• Node.js 18+ and npm
• Redis server (optional - fallback mode available)
• API keys for external services (optional - demo mode available)

Environment Configuration

The system works in demo mode by default with fallback implementations. For full functionality, configure these services:

# Authentication (Stytch)
NEXT_PUBLIC_STYTCH_PROJECT_ID=your_project_id
NEXT_PUBLIC_STYTCH_PUBLIC_TOKEN=your_public_token
STYTCH_SECRET=your_secret_key

# Context Engineering (Senso)
SENSO_API_KEY=your_senso_api_key
SENSO_ORG_ID=your_org_id

# Evaluation (HoneyHive)
HONEYHIVE_API_KEY=your_honeyhive_api_key

# Publishing (Qodo)
QODO_API_KEY=your_qodo_api_key

# Data Sources
BRIGHT_DATA_API_KEY=your_bright_data_key
APIFY_API_TOKEN=your_apify_token

# Vector Store (Redis)
REDIS_URL=redis://localhost:6379

# LLM APIs
OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key

Running the Demo

1. Development Mode:

   npm run dev

2. Production Build:

   npm run build
   npm start

3. Type Checking:

   npm run type-check

Demo Features

• Fallback Mode: All components work without external APIs
• Real-time Visualization: Live agent workflow display
• Interactive Demo: Click "Start Demo" to see the full pipeline
• API Testing: Built-in endpoints for testing integrations

🏗️ Architecture

Core Components

• Agent Orchestrator - Coordinates multi-agent workflow
• MCP Integrations - Bright Data, Apify, Senso integrations
• Context Engineering - PDF → chunks → schema → retrieval pipeline
• A2A Loops - Self-improvement through evaluation feedback
• Demo Interface - Real-time visualization of agent activity

Key Features

• Context Engineering: Transforms messy PDFs into structured, retrievable context
• A2A Communication: Agents collaborate and self-improve through feedback loops
• MCP Integration: Uses Model Context Protocol for tool integration
• Real-time Demo: Live visualization of agent workflow and decision-making

🎬 Demo Flow (3 minutes)

1. Start Demo - Click "Start Demo" to begin the workflow
2. Watch Agents Work - See each agent complete its specialized task
3. A2A Evaluation - Observe the self-improvement loop in action
4. Final Result - View the generated proposal and microsite

🛠️ Technical Implementation

Agent Types

• rfp-monitor - Discovers and monitors RFP portals
• pdf-processor - Extracts and chunks PDF content
• context-engineer - Normalizes data into structured schemas
• proposal-writer - Generates proposals using retrieved context
• evaluator - Evaluates proposals and generates improvement feedback
• deployer - Deploys final microsite

MCP Tools

• Bright Data MCP: Web scraping and data extraction
• Apify MCP: PDF processing and content extraction
• Senso MCP: Data normalization and schema mapping

Context Engineering Pipeline

1. PDF extraction and chunking
2. Requirement and criteria extraction
3. Schema normalization and validation
4. Vector embedding generation
5. Knowledge graph construction
6. Retrieval index building

🎯 Competition Focus

This demo is designed for the Context Engineering AI Agent Competition and showcases:

• Advanced Reasoning: Multi-step problem solving across agents
• Tool Integration: 3+ sponsor tools (Bright Data, Apify, Senso)
• A2A Communication: Self-improving agent loops
• Context Engineering: Sophisticated data transformation pipeline
• Live Demo: Real-time visualization of agent behavior

📁 Project Structure

src/
├── components/
│   └── DemoInterface.tsx      # Main demo interface
├── lib/
│   ├── agents/
│   │   └── orchestrator.ts    # Agent coordination
│   ├── mcp/
│   │   ├── bright-data.ts     # Bright Data MCP
│   │   ├── apify.ts          # Apify MCP
│   │   └── senso.ts          # Senso MCP
│   └── retrieval/
│       └── redis-llamaindex.ts # Vector store integration
├── types/
│   ├── rfp.ts                # RFP data types
│   └── agent.ts              # Agent communication types
└── app/
    └── page.tsx              # Main page

🔧 Configuration

Set up environment variables for MCP servers and API keys:

# MCP Server URLs
BRIGHT_DATA_MCP_URL=http://localhost:3001
APIFY_MCP_URL=http://localhost:3002
SENSO_MCP_URL=http://localhost:3003

# Redis Configuration
REDIS_URL=redis://localhost:6379

# API Keys
OPENAI_API_KEY=your_key_here
ANTHROPIC_API_KEY=your_key_here

🎪 Demo Script

1. Introduction (30s)

   • "This is Auto RFP, an AI agent system that processes RFPs end-to-end"
   • "It uses Context Engineering to transform messy PDFs into structured data"
   • "And A2A communication for self-improvement"

2. Live Demo (2m)

   • Click "Start Demo"
   • Show agent workflow progression
   • Highlight A2A evaluation loop
   • Display final proposal result

3. Wrap-up (30s)

   • "The system demonstrates advanced agent reasoning"
   • "Uses 3+ sponsor tools with MCP integration"
   • "Shows real-time A2A communication and self-improvement"

🏆 Competition Criteria

• Idea (25%): Unique value proposition with advanced reasoning
• Technical Implementation (25%): Novel use of tools and agent architecture
• Tool Use (25%): Integration of 3+ sponsor tools
• Presentation (25%): Clear 3-minute demo showcasing agent impact

📝 License

MIT License - Built for Context Engineering AI Agent Competition