CrewAI MCP Demo

This repository demonstrates different approaches for using CrewAI with MCP (Model Context Protocol) to interact with external tools and services through a standardized protocol.

🌟 Featured Examples

🏭 Enterprise Regulatory Intelligence

📍 Location: scaffolding_approach_examples/snowflake_mcp_demo/

A production-ready 3-agent system that combines CrewAI’s multi-agent orchestration with Snowflake’s enterprise data capabilities to build regulatory intelligence systems.

Key Features:

• 🤖 3-Agent Coordination: Regulatory Intelligence + SEC Filing Analysis + Market News
• 📊 Enterprise Scale: Semantic search across 372,299 SEC filings via Snowflake MCP
• 🎯 Professional Output: Investment-grade regulatory impact reports
• 🚀 Production Ready: Complete deployment and configuration examples

→ View Detailed Tutorial

---

🚀 Two Integration Approaches

This repository demonstrates two different ways to integrate CrewAI with MCP:

1. 🏗️ Scaffolding Approach (Recommended for Production)

• Uses CrewAI’s project scaffolding with @CrewBase decorator
• Structured project with configuration files
• Built-in get_mcp_tools() method (CrewAI 0.134.0+)
• Better for complex, maintainable applications

Examples:

• Snowflake Enterprise Demo - Regulatory intelligence system
• Mathematician Project - Math operations via local StdIO server
• Context7 Integration - Document search via streamable-HTTP

2. 📝 Script Approach (Great for Learning/Prototyping)

• Standalone Python scripts using MCPServerAdapter directly
• Quick to set up and experiment with
• Perfect for understanding MCP concepts

Examples:

• stdio_client_demo.py - Math operations
• sse_client_demo.py - Cloudflare docs search
• streamable_http_client_demo.py - Simple greeting
• multiple_servers_client_demo.py - Combined functionality

---

🔧 Transport Mechanisms

Streamable HTTP ⭐ (Recommended)

• Use case: Remote hosted servers (Context7, cloud services)
• Benefits: Simple setup, bidirectional communication, web-friendly
• Example: Snowflake MCP, Context7 API

Server-Sent Events (SSE)

• Use case: Real-time server-to-client updates
• Benefits: Good for streaming responses, widely supported
• Example: Cloudflare docs, real-time data feeds

Standard Input/Output (StdIO)

• Use case: Local development, testing, process-based servers
• Benefits: Simple local setup, no network required
• Example: Local math server, file processing tools

---